# EvalMap · changelog: what changed Source page: https://evalmap.ai/changelog/ # What changed. Every change to this site since it went up — 477 of them over 42 days, newest first. The list is taken from the history of the work itself rather than written alongside it, because a changelog kept separately drifts from the thing it describes within a week. Each line says what it touched: the numbers are data.json, the file the chart reads. The change that published this page is not on it yet — it appears with the next build. [← back to the atlas](https://evalmap.ai/) · [sources and licences](https://evalmap.ai/licenses/) ## 9 November 2026 - Release tracker: record the rerun HF sweep and the org list in the protocol the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: rowcheck sweeps HF repos reopened in the window the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: 9 November journal records the repeat sweep that closes the 08:52:41–08:53:20Z gap (0 new rows) the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 9 November journal completed; PROTOCOL §6 states how rowcheck treats HF errors and Wayback the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release protocol: rowcheck.py checks every grok row against the maker's page (§6), with fixtures; snowflake.py exposes when() for it the tools - Release tracker: 9 November rumor pass; Frantic Ferret added, six cards noted, seal 5–11 Nov the notes, the numbers, the pages, the raw extracts, the tools ## 8 November 2026 - Regenerate the changelog the pages, the raw extracts - Release protocol: grok signs in through grok.com OAuth with the xAI key variable unset; xAI handle @SpaceXSI the tools - Release tracker: 8 November rumor pass; Qoduck added, ERNIE 5.5 Flash and Hy4 weights raised to high, seal 5–11 Nov the notes, the numbers, the pages, the raw extracts, the tools ## 7 November 2026 - Regenerate the changelog the pages, the raw extracts - Jeep: g1 by Qoduck (Qwen4 Flash/Plus/Max?) (xhigh) (owner-supplied, 7 November) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: 7 November rumor pass; 8 releases added, Muse Realtime Avatar shipped, Inkling 2 Small weights and ERNIE 5.5 Flash added as cards, seal 5–11 Nov the notes, the numbers, the pages, the raw extracts, the tools ## 6 November 2026 - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Mercury 3 (high), 77.0% [72.6, 81.3], strict 41 (OpenRouter, Inception) the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Jeep: Mercury 3 g2 detail credits the effort to the owner, not the session log the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Jeep: g2 by Mercury 3 (high) (owner-supplied, 6 November) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Jeep: g1 by Devin SWE-2 Flash (max) (owner-supplied, 6 November) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Devin SWE-2 Flash (max) at 45 minutes, 83.7% [80.0, 87.5], strict 60 (9 re-runs) the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Devin SWE-2 Flash (max), 129/129 on 6.7 in a bubblewrap sandbox, 80.9% [76.2, 85.5] the numbers, the pages, the raw extracts ## 5 November 2026 - EvalMapBench audit: a partial point's SVG label and key number are the unranked "—", its key entry reads "(partial, unranked)" the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: label the Hy4 Flash card ‘Hy4 Flash announced’, like the other announced cards the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 5 November rumor pass; 7 releases added, Mistral Large 4 weights, Kimi K3.1 Pro weights and GPT Image 2.6 shipped, Hy4 weights and Hy4 Flash added as cards, close week 29 Oct–4 Nov, seal 5–11 Nov the notes, the numbers, the pages, the raw extracts, the tools ## 4 November 2026 - Regenerate the changelog the pages, the raw extracts - Atlas: AA snapshot of 4 November 2026 (Intelligence Index v5, 88 configurations) the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Hy4 (high), page-only row (129/129) through OpenRouter's Tencent endpoint the numbers, the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Releases: Hy4 weights still not out at 08:14 UTC on 4 November (Hugging Face, ModelScope, GitHub checked) the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Releases: Hy4 (4 Nov), the general release of Hy4 preview, with the cutoff moved to 4 November for that row only the numbers, the pages, the raw extracts, the tools ## 3 November 2026 - Regenerate the changelog the pages, the raw extracts - Releases: LongCat-2.5 (2 Nov), the general release of LongCat-2.5-Preview, with the cutoff moved to 3 November for that row only the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Jeep: game31 by LongCat 2.5 (max) (owner-supplied, 3 November) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add LongCat 2.5 to EvalMapBench the numbers, the pages, the raw extracts ## 29 October 2026 - Regenerate the changelog the pages, the raw extracts - Release tracker: the 29 October second-pass journal says the open week was reprinted for the GPT Image 2.6 card and notes where the §0 grep list was kept the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 29 October second pass; no new releases, GPT Image 2.6 added as an image card, notes added to DeepSeek V4.1 Pro, OpenAI's next model and Mistral Large 4 weights the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Odds: a settled question keeps its pointer after its card is counted and removed; odds.json as on 28 October the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: the 29 October journal says leftOut does name Granite Speech 5.0 2B and Raon-1-mini-preview the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 29 October pass; close week 22–28 Oct, seal 29 Oct–4 Nov the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: 28 October rumor pass; 7 releases added, Genie 4 added as a later card, Gemini 4 Carbon announced, week 22–28 Oct rechecked the notes, the numbers, the pages, the raw extracts, the tools ## 27 October 2026 - Regenerate the changelog the pages, the raw extracts - Jeep: the GPT-6.1 Terra card names game1 by its archive and project, not a folder the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Jeep: game1 by GPT-6.1 Terra (max) (owner-supplied, 27 October) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: the 27 October journal keeps Fern 0.1 on the watch list and names what stayed out of leftOut the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: the 27 October note and journal count Mirendil Fern 0.1 as the pass's new card the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: Mirendil Fern 0.1 announced 26 Oct, added as a later card the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 27 October rumor pass; GPT-6.1 Terra and Ling-3.1-flash-1019 shipped, 4 releases added, week 22–28 Oct rechecked the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench: GPT-6.1 Terra (max), page-only row (129/129) with its 45-minute re-runs the numbers, the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - EvalMapBench: GPT-6.1 Terra (high), page-only row (129/129) the numbers, the pages, the raw extracts ## 25 October 2026 - Regenerate the changelog the pages, the raw extracts - Release tracker: record the 25 October pass's deploy version and recheck the evening of 24 October the raw extracts - Regenerate the changelog the pages, the raw extracts - Release tracker: 25 October rumor pass; Nemotron 3.5 Bolt E4B and Kimi K3.1 Pro shipped, 9 releases added, week 22–28 Oct rechecked the numbers, the pages, the raw extracts, the tools ## 24 October 2026 - Regenerate the changelog the pages, the raw extracts - EvalMapBench: K3.1 Pro note says the default was high by 24 October the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - EvalMapBench: note that Kimi K3.1 Pro ran at max while Kimi's documented default moved to high the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Kimi K3.1 Pro (max), page-only row (129/129) the numbers, the pages, the raw extracts ## 23 October 2026 - Regenerate the changelog the pages, the raw extracts - Jeep: game23 by Kimi K3.1 Pro (max) (owner-supplied, 23 October) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: 23 October rumor pass; MiniMax M3.1 Flash weights shipped, week 22–28 Oct rechecked the numbers, the pages, the raw extracts, the tools ## 22 October 2026 - Release tracker: 21 and 22 October rumor passes; close week 15–21 Oct, seal 22–28 Oct the numbers, the pages, the raw extracts, the tools ## 21 October 2026 - EvalMapBench: Ornith 1.5 122B (high), page-only row (129/129) the numbers, the pages, the raw extracts ## 19 October 2026 - Regenerate the changelog the pages, the raw extracts - EvalMapBench: MiniMax M3.1 Flash, page-only row (129/129) the numbers, the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Jeep: game19 second build by Kimi K2.7 Code (owner-supplied, 19 October) the jeep builds, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Jeep: game17 by MiniMax M3.1 Flash (owner-supplied, 19 October) the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Atlas: DeepSeek V4.1 Flash 1017 point and jeep link the jeep builds, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Release tracker: 19 October rumor pass (with pending 16-18 October updates) the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools ## 17 October 2026 - EvalMapBench: DeepSeek V4.1 Flash 1017 (max) through Terminus-2 on DeepSeek's API the pages, the raw extracts ## 16 October 2026 - EvalMapBench: Pareto 26.10 through Terminus-2 on Unbiased's API the pages, the raw extracts - EvalMapBench: Nemotron 3.5 Super VL through Terminus-2 on paid OpenRouter the pages, the raw extracts ## 15 October 2026 - EvalMapBench: Ultramarine 1 Flash (max) through Terminus-2 on Paper Instruments' API the pages, the raw extracts - Jeep: Nemotron 3.5 Super VL, owner-supplied game9, failed (crashes on the first frame); rebuilt site the jeep builds, the numbers, the pages, the raw extracts, the tools - Releases 15 Oct: Nemotron 3.5 Bolt 74B (card added, then released) and Nemotron 3.5 Super VL the raw extracts - Snapshots of 15 October: AA through the Data API, fresh open feeds; method-text counts and the Pareto front follow the data the raw extracts, the tools - Releases: 15 Oct rumors pass — Step 5 Preview weights, Ling-3.1-flash weights, Grok Voice 3.0; forecast week 15–21 Oct the numbers, the pages, the raw extracts, the tools - EvalMapBench: note that Gemini 4 Argon (max) barely used its 1M output budget the pages, the raw extracts, the tools - EvalMapBench: Gemini 4 Argon (max) through cli2027 the numbers, the pages, the raw extracts - Jeep: Gemini 4 Argon (max, cli2027), owner-supplied game6 the jeep builds, the pages, the raw extracts, the tools ## 14 October 2026 - validate: every jeep shotNote must reach the home banner or be listed in OFF_BANNER the tools - Home banner: show owner-supplied jeep builds with race/countdown frames the pages, the tools - Jeep: Gemini 4 Argon (max, 1M output), owner-supplied game5 the jeep builds, the pages, the raw extracts, the tools - Odds: 14 Oct (no new maker dates; Muse Spark next and Grok 4.8 follow markets down) the pages, the raw extracts - Ledger record: score only the cards named for the week; a further-out card that shipped is said apart the pages, the tools - Ledger: close the week of 4-10 Oct (1 of 4 named shipped: Claude Haiku 5.5) the pages, the raw extracts - EvalMapBench 6.7: all 43 tasks ranked the numbers, the pages, the raw extracts, the tools - Releases: odds table links only to cards still shown the pages, the tools ## 13 October 2026 - Home banner: EvalMapBench ranks every full row, priced or not the pages, the tools - Strip: shipped cards (Argon, Watermelon) leave the expected side the pages, the tools - AA snapshot 13 Oct 2026: Muse Wonder in; Ling-3.0-flash-VL kept without AA price and speed the numbers, the pages, the raw extracts, the tools - Jeep: Devin SWE-2 (max), owner-supplied game4 the jeep builds, the pages, the raw extracts, the tools - Releases: pass 2026-10-13 (Muse Wonder, Gemini 4 Argon, MiniCPM-V 4.7, LFM2.5-24B-A2B) the numbers, the pages, the raw extracts, the tools - Jeep: Gemini 4 Argon tile captured natively on the RTX 5090 the jeep builds, the pages, the raw extracts, the tools - Jeep: Gemini 4 Argon (high), owner-supplied game3 the jeep builds, the pages, the raw extracts, the tools - EvalMapBench 6.6: Devin SWE-2 (max) complete, ranked the pages, the raw extracts, the tools - EvalMapBench point key: Muse Wonder Full keeps #5, no per-token price so no point the pages, the raw extracts, the tools - EvalMapBench: rank full rows without a price (Muse Wonder Full) the pages, the raw extracts, the tools - EvalMapBench 6.6: Muse Wonder Full (max) complete, ranked the pages, the raw extracts, the tools - EvalMapBench 6.6: all 40 tasks ranked the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools ## 12 October 2026 - EvalMapBench matrix: a column with the number of tasks each row solved the pages, the tools - EvalMapBench matrix: rows sorted by tasks solved, most first the pages, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench: Claude Fable 5.1 (high) through Claude Code, on all forty tasks the numbers, the pages, the raw extracts - EvalMapBench matrix: every row on the four new tasks (Opus 5.5 xhigh, Gemini 3.8 Flash and Kimi's re-run last) the pages, the raw extracts - EvalMapBench matrix: Sonnet 5.5 xhigh and Qwen3.8 Max Prime on the four new tasks the pages, the raw extracts - EvalMapBench matrix: Haiku 5.5, GLM-5.3-Prime and Qwen3.8 27B on the four new tasks the pages, the raw extracts - EvalMapBench matrix: the four new tasks on most rows so far; a dot marks cells run through Claude Code or Codex the numbers, the pages, the raw extracts, the tools - Releases 12 Oct: cutoff and forecast check to 12 October; Watermelon market back to a 15 Oct median the numbers, the pages, the raw extracts, the tools ## 11 October 2026 - Releases 11 Oct: Pixel Canary is Callstack's Apex; Qwen teaser, MiniCPM-V 4.7 loader and Watermelon market on their cards the numbers, the pages, the raw extracts, the tools - Home: What's new expanded on load; audit expects it open the pages, the tools - EvalMapBench matrix: models as rows, tasks as columns the pages, the tools - EvalMapBench matrix: compact squares (green solved, red not solved, grey not run) the pages, the tools - EvalMapBench: per-task matrix of every model on every task the pages, the raw extracts, the tools - Home UI: compact first screen, collapsible banners, readable EvalMapBench, two-model comparison the notes, the pages, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench chart: plot output tokens per trial instead of minutes the pages, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench star: price counts half as much as time and detail share the pages, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench 6.5: Muse Spark 1.3 Contributor at effort max joins the pages, the raw extracts - EvalMapBench star: leave the 0% baseline out of the frontier and its scales the pages, the tools - EvalMapBench: star the frontier model nearest the cheap, fast, highest-share corner; partial rows out of the frontier the pages, the tools - EvalMapBench 3D chart: star only the best model on the Pareto frontier (highest detail share) the pages, the tools - EvalMapBench 3D chart: numbered circles replace dots and callout boxes; Pareto star in key and table the pages, the tools - EvalMapBench 3D chart: perspective with drop lines and floor shadows instead of the isometric view; flat Price/Time presets unchanged the pages, the tools - Regenerate the changelog the pages, the raw extracts - EvalMapBench 6.5: ten new tasks, every row re-measured on thirty-six the numbers, the pages, the raw extracts - EvalMapBench: the detail chart as a turnable 3D cube of share, price and time; Price/Time become camera presets the pages, the tools - EvalMapBench: Price/Time toggle on the chart, median minutes per trial on a log axis the pages, the tools - EvalMapBench 6.5: ten new tasks, every row re-measured on thirty-six the numbers, the pages, the raw extracts, the tools ## 10 October 2026 - EvalMapBench 6.5: add 'The middle is a tie' limitation the tools - Regenerate the changelog the pages, the raw extracts - Generate SEO/LLM metadata: Dataset, ItemList and FAQ JSON-LD, llms-full.txt, markdown pages the pages, the tools - EvalMapBench 6.4: three new tasks, Qwen3.8 Max Prime high, Qwen3.8 27B on Cerebras (#13) the numbers, the pages, the raw extracts - Snapshots of 10 October: AA 196 configurations, fresh open feeds, odds day, releases through 10 Oct the notes, the numbers, the pages, the raw extracts, the tools ## 9 October 2026 - EvalMapBench: pointing at a key row lights its point on the chart the pages, the tools - EvalMapBench 6.3: Gemini 3.8 Flash at 61 of 69 trials the numbers, the pages, the raw extracts - EvalMapBench 6.3: Grok 4.7 high; Gemini 3.8 Flash at 56 of 69 trials the numbers, the pages, the raw extracts - EvalMapBench: chart axes fit the data, points coloured by maker with a legend the pages, the tools - EvalMapBench 6.3: Qwen3.8 Max Prime at 45 minutes; Gemini 3.8 Flash at 52 of 69 trials the numbers, the pages, the raw extracts - EvalMapBench 6.3: Grok 4.7 xhigh; Gemini 3.8 Flash at 47 of 69 trials the numbers, the pages, the raw extracts - EvalMapBench 6.3: Step 5 Preview at 45 minutes; Gemini 3.8 Flash at 32 of 69 trials the numbers, the pages, the raw extracts - EvalMapBench 6.3: Gemini 3.8 Flash, published while its run is still going the numbers, the pages, the raw extracts, the tools - EvalMapBench: results table fits the page width without scrolling (#12) the pages, the tools - EvalMapBench: ship table sorting as sort.js, the CSP blocks inline scripts (#11) the pages, the tools - EvalMapBench 6.3: scored beyond a gpt-oss-120b baseline the numbers, the pages, the raw extracts, the tools - EvalMapBench: tidier tier table, click-to-sort headers on both tables (#10) the pages, the tools - EvalMapBench 6.2: 45-minute figures for Sonnet 5.5 xhigh and high and Haiku 5.5 xhigh the pages, the raw extracts - EvalMapBench 6.2: Step 5 Preview joins, first 45-minute re-runs land the numbers, the pages, the raw extracts - EvalMapBench: tidier results table (notes under numbers, right-aligned headers) the pages, the tools - EvalMapBench 6.2: Claude Mythos 5.1 joins, tokens and 45-minute columns the numbers, the pages, the raw extracts, the tools - EvalMapBench 6.1: three planning tasks join the twenty, every row re-measured on them the numbers, the pages, the raw extracts, the tools - Read the EvalMapBench version and task count from the export the notes, the tools - EvalMapBench 6.0: Kimi K3 at max on Moonshot's own endpoint the numbers, the pages, the raw extracts - EvalMapBench 6.0: GLM-5.3-FlashX the pages, the raw extracts - EvalMapBench 6.0: GLM-5.3-Prime the pages, the raw extracts ## 8 October 2026 - EvalMapBench 6.0: GPT-6.1 Sol xhigh on OpenAI's ultrafast tier through Codex the pages, the raw extracts - EvalMapBench 6.0: GPT-6.1 Sol high on OpenAI's ultrafast tier through Codex the pages, the raw extracts - AA snapshot of 8 October: 187 configurations, Claude Haiku 5.5, and CritPt as a metric the notes, the numbers, the pages, the raw extracts, the tools - EvalMapBench 6.0: DeepSeek V4.1 Flash at max, the default run becomes its own high row the numbers, the pages, the raw extracts - Odds: the 8 October reading, eleven open questions to 31 Oct, 30 Nov and 31 Dec the pages, the raw extracts - Ledger: close the seven weeks that ended by 7 October the notes, the pages, the raw extracts, the tools - BridgeBench: re-read on 8 Oct with a new Onion key; 27 models, values unchanged the numbers, the pages, the raw extracts - Method text: the text arena reaches 71 configurations, not 68 the pages - LMArena: full refresh of all ten feeds (8 Oct) the numbers, the pages, the raw extracts - EvalMapBench 6.0: Claude Haiku 5.5 at xhigh, on the bench page until the atlas has it the pages, the raw extracts - EvalMapBench 6.0: DeepSeek V4.1 Flash ran at the API default, which is the high level the pages, the raw extracts - EvalMapBench 6.0: eleven rows, Anthropic and OpenAI efforts side by side the notes, the numbers, the pages, the raw extracts, the tools - Publish game22: GLM-5.3-Prime Jeep build, owner-supplied and unscored the jeep builds, the pages, the raw extracts, the tools - Vals: read the boards from the new page layout, refresh all six the numbers, the pages, the raw extracts, the tools - 8 October reading: forecast 8–14 Oct, open sources refreshed the notes, the numbers, the pages, the raw extracts, the tools - Banner: admit builds whose frame is from the RTX 5090 measured drive the pages, the tools - game21 card: say where the rainbow ground comes from the pages, the raw extracts - EvalMapBench page: a compact design the pages, the tools - Score game21 Step 5 Preview (high): 28/100, rubric v4, drives the jeep builds, the pages, the raw extracts, the tools - Publish game21: Step 5 Preview (high) Jeep build, owner-supplied and unscored the jeep builds, the pages, the raw extracts, the tools - Publish Claude Haiku 5.5 Jeep build and release, GPT-6 October in ChatGPT the jeep builds, the numbers, the pages, the raw extracts, the tools - EvalMapBench 6.0: allow rows the atlas has no entry for yet the tools - Bench-page audit: admit the disclosed Cloudflare page counter, as the main audit does the tools - Publish EvalMapBench 6.0 and retire the earlier tracks the notes, the numbers, the pages, the raw extracts, the tools ## 7 October 2026 - Publish Mythos 5.1 and Fable 5.1 xhigh Jeep builds and ribbon update the jeep builds, the pages, the raw extracts, the tools - Publish reviewed Bench UI, Mistral Large 4 and Nano Banana 2.1 the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools ## 6 October 2026 - feat: publish EvalMapBench aggregates and dated AA agent steps the notes, the numbers, the pages, the raw extracts, the tools - Releases: Banbury Road's Kardashev-0.7, announced 5 Oct, API beta waitlist, no date the pages, the raw extracts - Releases: Reflection AI's Beam, announced 5 Oct, weights promised later in October the pages, the raw extracts - Audit: expect the 6 Oct API snapshot (181 rows, SWE for all, 41 on the frontier); unplotted-model check on GPQA the tools - Take the 6 October AA snapshot from AA's official Data API (181 rows, SWE domain for all); steps for 29 of 31 agents; attribution and dates (deployed as e2200f6f) the notes, the numbers, the pages, the raw extracts, the tools - Jeep: publish Kolibri 1 (oc2) as a failed build, metadata only (deployed 5 Oct as d1f3f301) the jeep builds, the pages, the raw extracts, the tools ## 5 October 2026 - Re-read BridgeBench on 5 Oct (data unchanged); rebuild; dates the numbers, the pages, the raw extracts - Header: build keeps the agent steps date beside the AA snapshot date the tools - Y axis takes any metric; add AA Coding Agents mean steps (18 rows, read 5 Oct 2026); update dates; carry the uncommitted 4-5 Oct working tree (releases, EvalMapBench v5, jeep, rumors) the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools ## 3 October 2026 - EvalMapBench v5: refresh Fable and GLM, publish partial Flash and Qwen Prime the numbers, the pages, the raw extracts, the tools ## 2 October 2026 - Jeep: publish Saltwind Rally from Grok 4.7 Fast xhigh the jeep builds, the pages, the raw extracts, the tools - Jeep bench: grok47-xhigh-oc2 failed the brief (main.js does not parse) the jeep builds, the pages, the raw extracts, the tools - Jeep bench: fledge-alpha-oc2 failed the brief (three.tsl.js 404); declare free Zen models in the harness the jeep builds, the pages, the raw extracts, the tools - Jeep bench: fledge-alpha-oc2 (unscored: three/tsl 404), bench fixes for opencode 2.0.14 the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Fetch BridgeBench through the Onion account API the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Bring the hand-written counts in line with the 2 October snapshot the notes, the pages, the tools - Regenerate the changelog the pages, the raw extracts - Refresh the open sources and Artificial Analysis to 2 October the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Redesign the releases page: an editorial hero and a timeline of what shipped the pages, the tools - Regenerate the changelog the pages, the raw extracts - Deploy to Cloudflare after a green check on main the notes - List Ideogram 4.5 and FLUX 3 Image on the releases page the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Leave merge commits out of the changelog the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Tell search engines what each page is: titles, structured data, a fetchable icon and llms.txt the jeep builds, the notes, the pages, the tools ## 1 October 2026 - Redesign the atlas and the jeep gallery, add a dark theme, and take the 1 October snapshots the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add Ling-3.1-flash's jeep build at 37, run on the bench at its highest effort the jeep builds, the pages, the raw extracts, the tools ## 30 September 2026 - Regenerate the changelog the pages, the raw extracts - Add GPT-6.1 Sol to EvalMapBench: 142 of 147 tasks, run as GPT-6 Sol was the numbers, the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Take the 30 September Artificial Analysis snapshot: GPT-6.1 Sol on the atlas, with its blitz and knowledge the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add the hard blitz (v5.1) and take the 30 September blitz export: Sonnet 5 out by age the numbers, the pages, the raw extracts, the tools ## 29 September 2026 - Regenerate the changelog the pages, the raw extracts - Add GPT-6.1 Sol's jeep build at 45, made in the owner's own Codex the day it came out the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add GPT-6.1 Sol, released at DevDay on 29 September, which settles the DevDay question the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Bring the releases to 29 September: Kling 4.0 Flash, Kling 4.0 announced, OpenAI drops GPT-6.1 Astra the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Refresh AA and BridgeBench, read Sonnet 5.5's SWE index off AA's chart, and say on the chart when a picked model is not on it the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Open each build straight from its thumbnail on the atlas banner the pages, the tools - Regenerate the changelog the pages, the raw extracts - Take the 29 September blitz export: Sonnet 5.5 in progress, seven rows on the atlas, five older ones out the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add Claude Sonnet 5.5: its release, five configurations from AA, its jeep build at 48, its knowledge the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools ## 28 September 2026 - Regenerate the changelog the pages, the raw extracts - Add the evening's releases: NaiveAI's Naive-N0.5-Flash and ElevenLabs' Eleven v4 the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Bring the releases to 28 September: MiMo-V2.6 MOPD weights, Grok 4.7 from 20 Sep, OpenAI's pause the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add Pixel Canary's jeep build: Badlands Rally, 12/100, breaks at the rim the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Put the EvalMapBench blitz on the atlas: 146 small pieces, five rows with 90% of a verdict the numbers, the pages, the raw extracts, the tools ## 27 September 2026 - Run Gemini 3.8 Flash on an eighth harness, Google's Antigravity CLI: 12, breaks the jeep builds, the pages, the raw extracts, the tools - Bring the releases to 27 September: MiniMax-M3.1-Flash-Preview, and a strip that keeps its days the numbers, the pages, the raw extracts, the tools - Rerun the three builds NanoGPT's limit and a stalled provider cut off: one drives the jeep builds, the pages, the raw extracts, the tools - Publish the 27 September reruns: thirteen builds, two that drive the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Re-export EvalMapBench: Grok 4.7 and Kimi K3 join, on 60 and 52 of the 150 tasks the numbers, the pages, the raw extracts - Turn the BridgeBench button to one shot against overall, as the owner asked the pages, the tools - Note OpenCode Go's data terms and request limits, and the first MiniMax Code run's use the tools - Publish MiniMax-M3.1-Flash-Preview's jeep build from MiniMax Code, held to 12 the jeep builds, the pages, the raw extracts, the tools - Add BridgeBench as a source, and turn the chart with three large buttons the numbers, the pages, the raw extracts, the tools - Add a seventh jeep bench harness, MiniMax Code the tools - Let the jeep box's models test in a browser and look up package files the tools - Regenerate the changelog the pages, the raw extracts - Give EvalMapBench a banner of its own, a preset and a way into the 3D chart the pages, the tools - Regenerate the changelog the pages, the raw extracts - Link the EvalMapBench page from everywhere a reader looks the numbers, the pages, the tools - Put the jeep box's Chromium on the PATH as chromium the tools - Regenerate the changelog the pages, the raw extracts - Take the 27 September Artificial Analysis snapshot: 160 new speeds, two now EvalMap's the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Re-measure EvalMapBench v3 after fixing six tasks and three grading faults the numbers, the pages, the raw extracts - Add Space Bunny Alpha's second jeep run, through OpenCode Zen: it fails at load the jeep builds, the pages, the raw extracts, the tools - Add a sixth jeep bench harness, Freebuff, and its first eight builds the jeep builds, the pages, the raw extracts, the tools ## 26 September 2026 - Regenerate the changelog the pages, the raw extracts - Add EvalMapBench: private tasks from one developer's own work, and a page of solved against price the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Link LongCat-2.5-Preview's release row to its jeep build the pages, the raw extracts - Regenerate the changelog the pages, the raw extracts - Add LongCat-2.5-Preview to the jeep bench: 32, on the day OpenCode made it free the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Score the jeep bench on rubric v4, and add five builds and three failed cards the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Bring the releases to 26 September: Pixel Canary, LongCat-2.5-Preview, Perceptron Mk1.5 the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Take the odds for 26 September: part of the way toward the market on Sonnet 5.5 the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add GPT-5.6 Luna and Qwen3.7 Plus, and failed cards for Inkling Small and Gemini 3.5 Flash the jeep builds, the pages, the raw extracts, the tools ## 25 September 2026 - Regenerate the changelog the pages, the raw extracts - Add Claude Opus 5, Gemini 3.8 Flash and GPT-5.6 Terra from the Command Code bench the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Put the odds on the atlas, under the jeep banner the pages, the tools - Note what Command Code charges and which models refuse --effort max the tools - Regenerate the changelog the pages, the raw extracts - Add four more Command Code jeep builds: GPT-5.6 Sol, Qwen3.7 Max, Inkling, Gemini 3.6 Flash the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog the pages, the raw extracts - Add Claude Sonnet 5 from the Command Code bench, and stop counting its cache twice the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Qwen3.5 397B's failed jeep run the pages, the raw extracts - Run Qwen3.5 397B A17B on the jeep bench: it failed the brief, a HUD over a white screen the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for three more failed jeep runs through NanoGPT the pages, the raw extracts - Run DeepSeek V4 Flash 0731, Ling-3.0-flash and Kimi K2.6 on the jeep bench: all three failed the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Hy3's failed jeep run the pages, the raw extracts - Run Hy3 on the jeep bench: it failed the brief, "pow is not defined" the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for MiMo-V2.5-Pro's failed jeep run the pages, the raw extracts - Run MiMo-V2.5-Pro on the jeep bench: it failed the brief, stuck on its loading screen the jeep builds, the pages, the raw extracts, the tools - Let the banner check skip a build whose frame carries a note, as the banner does the tools - Regenerate the changelog for Kimi K3's jeep run the pages, the raw extracts - Run Kimi K3 on the jeep bench through Command Code: 33, scored on its WebGL2 fallback the jeep builds, the pages, the raw extracts, the tools - Price Command Code runs at list price, and let the capture tools hide WebGPU the tools - Regenerate the changelog for three failed jeep runs through NanoGPT the pages, the raw extracts - Run MiniMax-M3, GLM-5.2 and Kimi K2.7 Code on the jeep bench: all three failed the brief the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for DeepSeek V4 Pro 0813's failed jeep run the pages, the raw extracts - Run DeepSeek V4 Pro 0813 on the jeep bench: it failed the brief, nothing is ever drawn the jeep builds, the pages, the raw extracts, the tools - Add a Command Code runner to the jeep bench, for the models the owner's subscription carries the tools - Add Antigravity and grok CLIs to the ledger, and Go's DeepSeek V4 Pro now that Global is on the tools - Note in the ledger that OpenCode Go works again, and that DeepSeek V4 Pro there wants the Global region the tools - Start a ledger of subscriptions and CLIs: what each has left and which to run a model on the tools - Let the jeep bench run models through NanoGPT, and take logins OpenCode 2 keeps in its database the tools - Let jeepshots.js press a start button that says "race", as jeepdrive and jeeppilot already do the tools - Regenerate the changelog for GLM-5.3-Prime's failed jeep run the pages, the raw extracts - Run GLM-5.3-Prime on the jeep bench: it failed the brief, only the HUD is drawn the jeep builds, the pages, the raw extracts, the tools - Refresh OpenCode's catalogue from models.dev before every jeep bench run the tools - Describe the AA fetch as it is in the README and the helper scripts, not as a hand save the notes, the tools - Say in aa_extract.py how the AA page is actually fetched: one curl request since 21 September the tools - Regenerate the changelog for the 25 September snapshot and releases the pages, the raw extracts - Bring the releases to 25 September: five new rows, the week of 25 Sep - 1 Oct, and today's odds the notes, the numbers, the pages, the raw extracts, the tools - Take the 25 September Artificial Analysis snapshot: three rows in, six out, 155 new speeds the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog for the MiMo UltraSpeed speed run the pages, the raw extracts - Time MiMo-V2.6-Pro-UltraSpeed: 344 tokens/s against 26 for standard Pro, noted on its card the numbers, the pages, the raw extracts, the tools - Regenerate the changelog for the Qwen3.8-Max-Prime speed run the pages, the raw extracts - Time Qwen3.8-Max-Prime: 109 tokens/s against 49 for 0902, on one client and endpoint the numbers, the pages, the raw extracts - Regenerate the changelog for the Qwen3.8-Max-Prime note on the atlas the pages, the raw extracts - Note Qwen3.8-Max-Prime on Qwen3.8 Max (0902) in the atlas, not as a point of its own the numbers, the pages, the raw extracts, the tools - Regenerate the changelog for Qwen3.8-Max-Prime's release row the pages, the raw extracts - Add Qwen3.8-Max-Prime to the releases: Qwen3.8-Max in Alibaba's fast mode the pages, the raw extracts - Regenerate the changelog for Qwen3.8 Max Prime's jeep build the pages, the raw extracts - Add Qwen3.8 Max Prime's jeep build: Canyon Ridge Rally, 43/100 the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for the image test the pages, the raw extracts - Add a ten-model image test to the small tests: exact text in three scripts, clocks, hands, a full glass the pages, the raw extracts - Regenerate the changelog for the English role-play run the pages, the raw extracts - Rerun the Aion 3.5 role-play in English, and show the Russian run beneath it the pages, the raw extracts, the tools ## 24 September 2026 - Regenerate the changelog for the small tests page the pages, the raw extracts - Publish two small tests run here: Aion 3.5 on a long role-play, Recraft V4.1 Flash on text the notes, the pages, the raw extracts, the tools - Regenerate the changelog for Ling-3.0-flash-Fin's failed jeep run the pages, the raw extracts - Run Ling-3.0-flash-Fin on the jeep bench: it wrote no file the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for the Ling-3.0-flash-VL build the pages, the raw extracts - Run Ling-3.0-flash-VL on the jeep bench: 22, a mountain lake without a jeep in view the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for the DeepSeek-V4-Flash-Vision-Exp build the pages, the raw extracts - Run DeepSeek-V4-Flash-Vision-Exp on the jeep bench: 37, on ground that reads as water the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Granite 4.2 8B's failed jeep run the pages, the raw extracts - Run Granite 4.2 8B on the jeep bench: it failed the brief, and its card says why the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Mercury 2.5's failed jeep run the pages, the raw extracts - Run Mercury 2.5 on the jeep bench: it failed the brief in 18 seconds, and its card says why the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Qwen3.8-Flash-Next's jeep build the pages, the raw extracts - Run Qwen3.8-Flash-Next on the jeep bench and publish its build: 39, drives the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for Solar Mini 4's failed jeep run the pages, the raw extracts - Run Solar Mini 4 on the jeep bench: it failed the brief, and its card says why the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for the failed jeep mark the pages, the raw extracts - Show a failed jeep build red and struck through on the release line and list the pages, the tools - Regenerate the changelog for the odds the pages, the raw extracts - Put a number on the forecast: the odds by the next three month-ends, beside the markets the pages, the raw extracts, the tools - Regenerate the changelog for the failed-build cards the pages, the raw extracts - Show builds that failed the brief: AliceAI-Foundation gets a struck-through jeep and the reason the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for the Meta recheck the pages, the raw extracts - Recheck Meta after the Connect keynote: no model shipped, Muse Realtime Avatar announced the pages, the raw extracts - Add EvalMapBench: thirteen models on one developer's own work, measured here the numbers, the pages, the raw extracts, the tools ## 23 September 2026 - Regenerate the changelog for the single pick line the pages, the raw extracts - Keep one line on the 3D chart, from the previous pick to the current one the pages, the tools - Regenerate the changelog for the selection trail the pages, the raw extracts - Join the models picked on the 3D chart with one trail, in the order picked the pages, the tools - Regenerate the changelog for Space Bunny Alpha's jeep build the pages, the raw extracts - Publish Space Bunny Alpha's jeep build from the OpenCode bench, run twice the jeep builds, the pages, the raw extracts, the tools - Regenerate the changelog for wheel zoom the pages, the raw extracts - Zoom the 3D chart with the mouse wheel, within the bounds of the + and − buttons the pages, the tools - Regenerate the changelog for Space Bunny Alpha the pages, the raw extracts - Add Space Bunny Alpha, a stealth model that appeared on OpenRouter and OpenCode on 23 September the pages, the raw extracts - Regenerate the changelog for the 23 September release update the pages, the raw extracts - Update the release list and forecast to 23 September the notes, the numbers, the pages, the raw extracts, the tools - Regenerate the changelog, which had fallen behind the history the pages, the raw extracts - Take the 23 September Artificial Analysis snapshot: new speeds, and 16 rows AA now marks deprecated the notes, the numbers, the pages, the raw extracts, the tools - Retake five jeep tiles with the pilot on the course, not the jeep off it the jeep builds, the notes, the tools - Score Claude Code and Codex builds with run.py, not a rubric v1 of their own the tools - Steer the jeep pilot along the course and re-judge the builds it could follow the notes, the pages, the raw extracts, the tools - Run every jeep-bench model in a container with no way out but its API and npm the notes, the tools - Hold a jeep build that cannot be driven to 20 or 12, after a minute of steering it on the graphics card the notes, the pages, the raw extracts, the tools - Score every jeep build on rubric v2 after a measured drive on a graphics card, and retake the gallery there the jeep builds, the notes, the pages, the raw extracts, the tools - Take Artificial Analysis's speed for all twelve GPT-6 Sol and GPT-6 Luna configurations the numbers, the pages, the raw extracts, the tools - Regenerate the changelog, which had fallen eight changes behind the history the pages, the raw extracts - Open the README with a compact index of its eleven sections the notes - Make the jeep rubric v2: score what plays on screen, and put the measured frame rate on the card the pages, the tools ## 22 September 2026 - Take Artificial Analysis's speed for Claude Opus 5.5 at xhigh, high, medium and low the notes, the numbers, the pages, the raw extracts, the tools - Put GPT-6 Sol and GPT-6 Luna on the chart, all twelve configurations from Artificial Analysis the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - Put Claude Opus 5.5 on the chart, and publish the jeep builds of Opus 5.5, GPT-6 Sol and GPT-6 Luna the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - List GPT-6 Sol and GPT-6 Luna, out on 22 September, and take them off the week's forecast the pages, the raw extracts - Move Claude Opus 5.5 from the forecast to the releases, out on 22 September the notes, the numbers, the pages, the raw extracts, the tools - Put the scoring commit into the changelog it left out the pages, the raw extracts - Score every build in the gallery, not only the ones the bench ran the notes, the pages, the raw extracts, the tools - Score every bench build from 0 to 100, where 100 is a game on the level of GTA VI the notes, the pages, the raw extracts, the tools - Run the jeep brief ourselves on an OpenCode bench, starting with GLM-5.3 at max effort the jeep builds, the notes, the pages, the raw extracts, the tools - Let a build's page say the language its words are actually in the jeep builds, the notes, the pages, the raw extracts, the tools - Keep the two presets that name a job, and publish the build MiMo-V2.6-Pro wrote the jeep builds, the notes, the pages, the raw extracts, the tools - Stop the checks running on a version of Node the runner is about to withdraw the notes, the pages, the raw extracts - Give every page the audit opens a timeout a machine without a graphics card can meet the pages, the raw extracts, the tools - Keep the forecast to its word, and publish every change this site has been through the notes, the pages, the raw extracts, the tools - Date every source by its own download, and stop a stale checkpoint setting a price the notes, the numbers, the pages, the raw extracts, the tools - Mark DeepSeek V4.1-Pro, Watermelon and Grok 4.8 as announced, and keep each announcement visible at any width the notes, the pages, the raw extracts, the tools - Add MiMo-V2.6's UltraSpeed mode and 9B distillation, and Apsara's image and speech announcements the notes, the pages, the raw extracts, the tools - Put OpenAI's launch on Tuesday, and add three models due later from a rumour check the notes, the pages, the raw extracts - Keep an announcement on the strip until the model ships the pages, the tools - Put Qwen4 on the release strip on the day it was announced the notes, the pages, the raw extracts, the tools - Correct the Apsara note after a grok check, and add two Qwen speech models it turned up the pages, the raw extracts - Put MiMo-V2.6-Pro on the chart, list the MiMo-V2.6 release, and record Qwen4 as announced the notes, the numbers, the pages, the raw extracts, the tools ## 21 September 2026 - Take Grok 4.7's speed and price from Artificial Analysis, and retire the stand-ins the notes, the numbers, the pages, the raw extracts, the tools - Move each jeep build down under its strip, and add GLM-5.3-FlashX and Fugu Max the jeep builds, the notes, the pages, the raw extracts, the tools - Count Cloudflare's own beacon apart in the strip check the tools - Add Vals AI as a source, and give every jeep build a strip of its own the jeep builds, the notes, the numbers, the pages, the raw extracts, the tools - Take Grok 4.7 onto the chart with its own speed, price and SWE index the notes, the numbers, the pages, the raw extracts, the tools - Rework the LMArena join and add five arena boards the notes, the numbers, the pages, the raw extracts, the tools - Count the WebDev arena at 49 on the licences page too the pages - Give Qwen3.8 Max (0902) its own arena entry, not the older checkpoint's the notes, the numbers, the pages, the tools - Put a model released on the snapshot's day on the Released line the notes, the pages, the tools - Show the jeep builds only here, and give the model out today a full chip the notes, the pages, the tools - Add Grok 4.7's jeep race, published only here the jeep builds, the pages, the raw extracts, the tools - Write down how to gather releases and rumours with the grok CLI the tools - Mark each release's place on the chart and its jeep build, and add Grok 4.7 the notes, the pages, the raw extracts, the tools - Set the releases against today, lay out the week, and ring new models on the chart the notes, the numbers, the pages, the raw extracts, the tools - List the month's model releases and this week's forecast on the site the notes, the pages, the raw extracts, the tools - Keep the month's model timeline and this week's forecast the raw extracts - Call two builds by the names their makers gave the models the notes, the pages, the raw extracts, the tools - Open the gallery with an index of every build the notes, the pages, the tools - Add the five jeep builds that were published as Workers the jeep builds, the notes, the pages, the raw extracts, the tools - Make the audits count the injected counter, as the last commit said the tools - Say that the zone's counter reaches the game pages too the pages, the tools - Bump the licences page to the same stylesheet token the pages - Take the frontier table off the page the pages, the tools - Host every jeep build here, in English, on a timeline that reads downward the jeep builds, the notes, the pages, the raw extracts, the tools - Run the ribbon newest first, and sign the banner thumbnails the notes, the pages, the tools - Divide the jeep rail by week, and say it is an order the pages, the tools - Show what 21 models made of the same jeep brief the jeep builds, the notes, the pages, the raw extracts, the tools - Add nine metrics and two price sources, all openly licensed the notes, the numbers, the pages, the raw extracts, the tools ## 20 September 2026 - Say that Cloudflare's edge adds a page counter, because it does the notes, the pages, the tools - State plainly what the site takes and collects: nothing the notes, the pages, the tools - Rebuild the page around the chart, on Fable's design review the notes, the pages, the tools - Drop the 2D charts and leave the cube as the only view the notes, the pages, the tools - Darken the markers above the price limit and publish a contact address the notes, the pages, the tools - Put the green reference plane back and let it move along axis Z the notes, the pages, the tools - Swap the X and Z metrics so the benchmark reads vertically the pages - Put output speed back on an axis and price back in the cloud the notes, the pages, the tools - State precisely which Artificial Analysis terms apply and why none of them cover this site the notes, the numbers, the pages, the tools - Label every value with its source and let the visitor choose it the notes, the numbers, the pages, the raw extracts, the tools - Publish complete EvalMap source and Cloudflare deployment configuration the notes, the numbers, the pages - Initialize EvalMap website repository the notes