Models

Frontier & open-weight models compared by capability — columns are flagship models, rows are the headline benchmarks that measure each one. Every number is the published figure, restated with a link to the source it came from · as of 2026-06-14. Curated matrix is 70d old — a refresh is due; live leaderboard moves still flow into Model changes below.

We cite the benchmark authorities rather than re-rank them — Epoch AI, LMArena and Artificial Analysis. ChangeRadar's job is the changes: what gets released, repriced, deprecated, or quietly shifts behind a stable model id.

Vendor champions · best flagship per vendor · GPQA Diamond

Google

Gemini 3.1 Pro

94.3GPQA Diamond

Compare ⇄

Moonshot AI reported

Kimi K3

93.5GPQA Diamond

Source ↗

Alibaba reported

Qwen3.8-Max

92.6GPQA Diamond

Source ↗

Zhipu AI (Z.ai) open

GLM-5.2

91.2GPQA Diamond

Compare ⇄

OpenAI reported

GPT-5.6 Sol

90.4BrowseComp

Source ↗

DeepSeek reported

DeepSeek-V4-Pro-Max

90.1GPQA Diamond

Source ↗

xAI reported

Grok 4.6

88.4LiveCodeBench

Source ↗

Anthropic reported

Claude Opus 5

84.1GPQA Diamond

Source ↗

Mistral AI open

Mistral Medium 3.5

77.6SWE-bench Verified

Compare ⇄

Meta open

Llama 4 Maverick

69.8GPQA Diamond

Compare ⇄

Each vendor's most recent publicly-available flagship, ordered by score (highest first) — Google · OpenAI · Anthropic · xAI highlighted. The weekly market-watch surfaces new releases automatically; one tagged reported is the latest release shown with vendor-reported scores (linked to source) until we independently cite it. Score = GPQA Diamond; every number links to its source.

Compare two models

Qwen3-235B-A22B-Instruct-2507 — AlibabaQwen3-235B-A22B-Thinking-2507 — AlibabaQwen3-Coder-480B-A35B-Instruct — AlibabaQwen3-Max — AlibabaQwen3.7-Max — AlibabaClaude Fable 5 — AnthropicClaude Haiku 4.5 — AnthropicClaude Mythos 5 — AnthropicClaude Opus 4.6 — AnthropicClaude Opus 4.7 — AnthropicClaude Opus 4.8 — AnthropicClaude Sonnet 4.6 — AnthropicDeepSeek-R1-0528 — DeepSeekDeepSeek-V3.1 — DeepSeekDeepSeek-V3.2 — DeepSeekDeepSeek-V3.2-Exp — DeepSeekDeepSeek-V3.2-Speciale — DeepSeekDeepSeek-V4-Pro — DeepSeekGemini 2.5 Pro — GoogleGemini 3 Flash — GoogleGemini 3 Pro — GoogleGemini 3.1 Pro — GoogleGemma 3 12B IT — GoogleGemma 3 1B IT — GoogleGemma 3 27B IT — GoogleGemma 3 4B IT — GoogleLlama 4 Behemoth — MetaLlama 4 Maverick — MetaLlama 4 Scout — MetaCodestral 25.01 — Mistral AIMagistral Medium 1.2 — Mistral AIMagistral Small 1.2 — Mistral AIMinistral 3 — Mistral AIMistral Large 3 — Mistral AIMistral Medium 3.5 — Mistral AIKimi K2 — Moonshot AIKimi K2 Thinking — Moonshot AIKimi K2.6 — Moonshot AIKimi K2.7-Code — Moonshot AIGPT-5 — OpenAIGPT-5.1 — OpenAIGPT-5.2 — OpenAIGPT-5.5 — OpenAIGrok 4 — xAIGrok 4 Fast — xAIGrok 4 Heavy — xAIGrok 4.1 — xAIGrok 4.1 Fast — xAIGrok 4.3 — xAIGLM-4.6 — Zhipu AI (Z.ai)GLM-4.6V — Zhipu AI (Z.ai)GLM-4.7 — Zhipu AI (Z.ai)GLM-5.2 — Zhipu AI (Z.ai) ⇄ Qwen3-235B-A22B-Instruct-2507 — AlibabaQwen3-235B-A22B-Thinking-2507 — AlibabaQwen3-Coder-480B-A35B-Instruct — AlibabaQwen3-Max — AlibabaQwen3.7-Max — AlibabaClaude Fable 5 — AnthropicClaude Haiku 4.5 — AnthropicClaude Mythos 5 — AnthropicClaude Opus 4.6 — AnthropicClaude Opus 4.7 — AnthropicClaude Opus 4.8 — AnthropicClaude Sonnet 4.6 — AnthropicDeepSeek-R1-0528 — DeepSeekDeepSeek-V3.1 — DeepSeekDeepSeek-V3.2 — DeepSeekDeepSeek-V3.2-Exp — DeepSeekDeepSeek-V3.2-Speciale — DeepSeekDeepSeek-V4-Pro — DeepSeekGemini 2.5 Pro — GoogleGemini 3 Flash — GoogleGemini 3 Pro — GoogleGemini 3.1 Pro — GoogleGemma 3 12B IT — GoogleGemma 3 1B IT — GoogleGemma 3 27B IT — GoogleGemma 3 4B IT — GoogleLlama 4 Behemoth — MetaLlama 4 Maverick — MetaLlama 4 Scout — MetaCodestral 25.01 — Mistral AIMagistral Medium 1.2 — Mistral AIMagistral Small 1.2 — Mistral AIMinistral 3 — Mistral AIMistral Large 3 — Mistral AIMistral Medium 3.5 — Mistral AIKimi K2 — Moonshot AIKimi K2 Thinking — Moonshot AIKimi K2.6 — Moonshot AIKimi K2.7-Code — Moonshot AIGPT-5 — OpenAIGPT-5.1 — OpenAIGPT-5.2 — OpenAIGPT-5.5 — OpenAIGrok 4 — xAIGrok 4 Fast — xAIGrok 4 Heavy — xAIGrok 4.1 — xAIGrok 4.1 Fast — xAIGrok 4.3 — xAIGLM-4.6 — Zhipu AI (Z.ai)GLM-4.6V — Zhipu AI (Z.ai)GLM-4.7 — Zhipu AI (Z.ai)GLM-5.2 — Zhipu AI (Z.ai)

2 wins · Claude Fable 5 1 wins · Gemini 3.1 Pro 0 ties

Agentic coding

Benchmark	Claude Fable 5	Gemini 3.1 Pro	Δ
SWE-bench Verified i% resolved (pass@1)	95	80.6	+14.4
SWE-bench Pro i% resolved (pass@1)	80.3	54.2	+26.1

3 coverage gaps — only one model reports these

• Terminal-Bench — Gemini 3.1 Pro 68.5
• LiveCodeBench — Gemini 3.1 Pro 2887
• FrontierCode — Claude Fable 5 29.3

Tool use & agents

5 coverage gaps — only one model reports these

• TAU-bench — Gemini 3.1 Pro 99.3
• OSWorld — Claude Fable 5 85
• BrowseComp — Gemini 3.1 Pro 85.9
• GDPval-AA — Claude Fable 5 1932
• MCP Atlas — Gemini 3.1 Pro 69.2

Science & reasoning

Benchmark	Claude Fable 5	Gemini 3.1 Pro	Δ
GPQA Diamond i% accuracy	92.6	94.3	-1.7

2 coverage gaps — only one model reports these

• Humanity's Last Exam — Gemini 3.1 Pro 44.4
• ARC-AGI-2 — Gemini 3.1 Pro 77.1

General knowledge

1 coverage gap — only one model reports these

• MMMLU — Gemini 3.1 Pro 92.6

Multimodal

2 coverage gaps — only one model reports these

• MMMU — Gemini 3.1 Pro 80.5
• GDP.pdf — Claude Fable 5 29.8

Long context

1 coverage gap — only one model reports these

• MRCR — Gemini 3.1 Pro 84.9

Shared benchmarks first (with Δ when both report the same scale); one-sided coverage collapses below. Pick any two models — or a champion above — and the URL becomes shareable.

All models · benchmark matrix

⌕

Vendor

AnthropicOpenAIGoogleMoonshot AIMetaxAIDeepSeekAlibabaZhipu AI (Z.ai)Mistral AI

Capability

Open weightsAgentic codingTool use & agentsMathScience & reasoningGeneral knowledgeMultimodalLong contextHuman preference

19 of 19 columns shown

reported columns (Claude Opus 5, GPT-5.6 Sol, Grok 4.6, Qwen3.8-Max, DeepSeek-V4-Pro-Max, Kimi K3) are auto-discovered by our weekly market-watch from each vendor's own reported numbers — not independently verified, and shown when a vendor ships a model newer than the hand-cited column beside it. Full claim sets are in Vendor-reported benchmarks below.

Agentic coding

Can the model fix real bugs, ship features, and operate a dev environment end-to-end as a coding agent — the single most-watched capability in 2026 vendor launches.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
SWE-bench Verified i % resolved (pass@1)	95	97	88.6	—	—	—	80.6	—	—	80.6	80.6	80.4	—	80.2	76.8	—	—	77.6	—
SWE-bench Pro i % resolved (pass@1)	80.3	79.2	69.2	77.8	58.6	—	54.2	—	—	—	—	60.6	67.7	58.6	—	62.1	—	—	—
Terminal-Bench i % solved (pass@1)	—	68.9	74.6	88	82.7	88.8	68.5	—	26	67.9	—	69.7	86.6	66.7	88.3	81	—	—	—
LiveCodeBench i % pass@1	—	—	—	—	—	—	2887	—	88.4	93.5	93.5	—	—	89.6	—	—	43.4	—	29.7

Tool use & agents

Beyond writing code: can the model select and chain tools, follow policy, drive a computer/browser, and complete long-horizon multi-step tasks.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
TAU-bench i % pass / pass^k	—	—	—	—	—	—	99.3	98	—	—	—	—	—	—	—	—	—	—	—
OSWorld i % success	85	70.6	83.4	—	78.7	—	—	—	—	—	—	—	86.1	73.1	—	—	—	—	—
BrowseComp i % accuracy	—	90.8	84.3	—	90.1	90.4	85.9	—	—	—	—	—	—	83.2	91.2	—	—	—	—

Math

Competition and research-level mathematical reasoning, increasingly reported on uncontaminated/post-cutoff problem sets.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
HMMT i % accuracy (pass@1)	—	—	—	—	—	—	—	—	—	—	—	97.1	—	92.7	—	—	—	—	—
MATH i % accuracy	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	61.2	—	89

Science & reasoning

Expert-level, Google-proof reasoning across the sciences and broad academia — the benchmarks vendors point to when claiming 'PhD-level' or 'frontier' reasoning.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
GPQA Diamond i % accuracy	92.6	84.1	93.6	—	93.6	—	94.3	90.1	—	90.1	90.1	92.4	92.6	90.5	93.5	91.2	69.8	—	42.4
Humanity's Last Exam i % accuracy	—	64.7	57.9	64.5	57.2	—	44.4	—	—	37.7	—	41.4	43.6	54	43.5	40.5	—	—	—
ARC-AGI-2 i % solved	—	90.4	—	—	85	—	77.1	—	—	—	—	—	—	—	—	—	—	—	—

General knowledge

Broad multi-subject factual and reasoning coverage; the classic 'how much does it know' bucket, now reported via the harder Pro variant since base MMLU is saturated.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
MMLU-Pro i % accuracy	—	—	—	—	—	—	—	—	—	87.5	87.5	—	—	—	—	—	80.5	—	67.5

Multimodal

Vision + language understanding and visual reasoning — how well the model interprets images, diagrams, charts and figures.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
MMMU i % accuracy	—	—	—	—	83.2	—	80.5	—	—	—	—	—	—	79.4	—	—	73.4	—	64.9

Long context

Retrieval and reasoning quality as context length grows into the hundreds-of-thousands / millions of tokens — beyond simple needle-in-a-haystack.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
MRCR i % accuracy	—	—	—	—	74	—	84.9	—	—	—	—	—	—	—	—	—	—	—	—

Human preference

Aggregate real-user preference from blind head-to-head comparisons — the closest thing to a 'do people actually like the answers' metric, and the one number vendors love to top.

Benchmark	Fable 5	Opus 5 reported	Opus 4.8	Mythos 5	GPT-5.5	GPT-5.6 Sol reported	Gemini 3.1 Pro	Grok 4.3	Grok 4.6 reported	DeepSeek-V4-Pro open	DeepSeek-V4-Pro-Max reported	Qwen3.7-Max	Qwen3.8-Max reported	Kimi K2.6 open	Kimi K3 reported	GLM-5.2 open	Llama 4 Maverick open	Mistral Medium 3.5 open	Gemma 3 27B open
LMArena Elo i Elo	—	—	—	1458	1474	—	—	—	—	—	—	—	—	1466	—	—	1417	—	1338

Showing 13 flagship models across the headline benchmarks; 53 models and 84 benchmarks tracked in total from 60 primary & aggregator sources. Claude Mythos 5 numbers are limited (access-restricted preview). Numbers are published facts, restated with a per-cell source link; vendor benchmark charts are linked to their source, not rehosted.
Sources include: 9to5Google (Gemini 3 Flash launch coverage) · AIFire (citing OpenAI GPT-5.2 release) · Artificial Analysis · BenchLM.ai (citing LMArena) · BinaryVerse AI (xAI official figures) · BuildFastWithAI (citing OpenAI GPT-5.5 launch) · Caylent · Codersera (reporting Moonshot's figures) · DataCamp · DataCamp (reproducing Meta's Llama 4 launch chart) · DeepSeek-AI (arXiv 2512.02556) · DeepSeek-AI (Hugging Face model card) · Google (official Gemini 2.5 launch blog) · Google (official Gemini 3 Flash launch blog) · Google (official Gemini 3 launch blog) · Google DeepMind (Gemini 3.1 Pro model card) · +36 more.

Vendor-reported benchmarks

Numbers as claimed by the vendor on their own model/system card — not independently verified and often measured with a favourable harness. We track each vendor's claims over time and link to the source; cross-check against the cited matrix above.

Claude Opus 5 Anthropic

Frontier-Bench v0.1 · High effort	43.3%
ARC-AGI-3 · High effort	30.2%
SWE-bench Verified	97.0%
SWE-bench Pro	79.2%
GPQA Diamond	84.1%
Terminal-Bench 2.0	68.9%
OSWorld 2.0	70.6%
Humanity's Last Exam · with tools	64.7%
BrowseComp · agentic search	90.8%
GDPval-AA v2 · knowledge work	1861 Elo
ARC-AGI-1 · Max reasoning effort	97.5%
ARC-AGI-2 · Semi-Private, Max reasoning effort	90.4%

vendor card ↗

GPT-5.6 Sol OpenAI

Terminal-Bench 2.1 · base Sol; Sol Ultra scored 91.9%	88.8%
Agents' Last Exam · max reasoning setting	53.6 score
ARC-AGI-3 · with retained reasoning and compaction enabled	38.3%
ARC-AGI-3 · official harness (standard)	13.3%
DeepSWE	72.7%
BrowseComp · base Sol; Sol Ultra scored 92.2%	90.4%
ExploitBench	73.5%
Capture-the-Flag · OpenAI's curated internal task set; tool-enabled harness	96.7%
HealthBench Professional	60.5 length-adjusted score

vendor card ↗

Grok 4.6 xAI

Artificial Analysis Intelligence Index · nine-benchmark aggregate	61 composite
CursorBench v3.2 · high thinking effort	69.9%
DeepSWE v1.1 · high thinking effort	65.9%
FrontierCode v1.1 Extended · extended split	61.3%
APEX-Agents · high thinking effort	57.5%
APEX-SWE · high thinking effort	56.4%
Terminal-Bench v3.0 · Grok Build harness	26%
GDPval-AA v2	1753 Elo
AA-Briefcase · long-horizon analyst work	1577 Elo
Harvey Legal Agent Benchmark · Vals scoring	15.8%
LiveCodeBench · xhigh thinking effort	88.4%
Next.js Evals	92%

vendor card ↗

Gemini 3.1 Pro Google

GPQA Diamond · Graduate-level science	94.3%
ARC-AGI-2 · Abstract reasoning	77.1%
SWE-Bench Verified · Agentic code evaluation	80.6%
Humanity's Last Exam · No tools	44.4%
LiveCodeBench Pro · Competitive programming	2887 Elo
Terminal-Bench 2.0 · Terminus-2 harness	68.5%
SciCode · Scientific research coding	59.0%
BrowseComp · Agentic task	85.9%
MMMLU · Multilingual	92.6%
MCP Atlas · Tool use coordination	69.2%
APEX-Agents · Agentic multi-step tasks	33.5%

vendor card ↗

Qwen3.8-Max Alibaba

GPQA Diamond · Vendor-run; independent replication 92.6% vs claimed 92.7%	92.6%
SWE-bench Pro · Alibaba's own corrected task set with Claude Code harness	67.7%
Terminal-Bench 2.1 · 5-hour timeout; official leaderboard caps at 83.8%	86.6%
PaperBench · 28-point jump from Qwen3.7-Max; reproduces research paper results	93.0%
IFBench · Instruction following benchmark	82.8%
OSWorld-Verified · Desktop OS interaction; computer-use agentic workflows	86.1%
Humanity's Last Exam · Weakest showing among four flagship comparisons	43.6%
MathVision · Multimodal benchmark; vision-based math tasks	95.2%

vendor card ↗

DeepSeek-V4-Pro-Max DeepSeek

SWE-bench Verified · Pass@1	80.6%
GPQA Diamond · Pass@1	90.1%
MMLU-Pro · EM	87.5%
LiveCodeBench	93.5 pass@1

vendor card ↗

Llama 4 Maverick Meta

MMLU Pro · 5-shot, CoT	80.5 %
GPQA Diamond · 0-shot, CoT, multiple generations averaged	69.8 %
LiveCodeBench · 10.01.2024-02.01.2025, multiple generations averaged	43.4 %
MMMU · Image reasoning	73.4 %
MathVista · Image reasoning	73.7 %
ChartQA · Image understanding	90.0 %
DocVQA · Image understanding, test set	94.4 %
Multilingual MMLU · Multilingual	84.6 %
MTOB Half Book · Long context, eng->kgv/kgv->eng	54.0 / 46.4 %
MTOB Full Book · Long context, eng->kgv/kgv->eng	50.8 / 46.7 %

vendor card ↗

Mistral Medium 3.5 Mistral AI

SWE-Bench Verified	77.6%
τ³-Telecom	91.4%

vendor card ↗

Kimi K3 Moonshot AI

GPQA Diamond	93.5%
SWE-Bench Verified	76.8%
Terminal-Bench 2.1	88.3
FrontierSWE	81.2 pass@1
Program Bench · raw pass rate	77.8%
SWE Marathon	42.0
DeepSWE · Kimi Code harness	67.5
BrowseComp · 300K context compaction	91.2
Humanity's Last Exam (HLE-Full) · text / text+tools	43.5 / 56.0
CritPt	23.4
AA-LCR · Long-context retrieval	74.7

vendor card ↗

GLM-5.2 Zhipu AI (Z.ai)

SWE-bench Pro	62.1%
Terminal-Bench 2.1	81.0%
GPQA Diamond	91.2%
AIME 2026	99.2%
HLE · with tools	54.7%

vendor card ↗

Model changes

New releases, deprecations, and benchmark score moves we’ve recorded — newest first.

• Vendor claim Meta Llama — reported benchmarks Aug 23, 2026

  Meta reported benchmarks updated

  Llama 4 Maverick: 10 benchmark claims (via web search)

• Vendor claim OpenAI — reported benchmarks Aug 22, 2026

  OpenAI reported benchmarks updated

  GPT-5.6 Sol: 9 benchmark claims (via web search)

• Vendor claim Moonshot Kimi — reported benchmarks Aug 22, 2026

  Moonshot AI reported benchmarks updated

  Kimi K3: 11 benchmark claims (via web search)

• Vendor claim Zhipu GLM — reported benchmarks Aug 22, 2026

  Zhipu AI (Z.ai) reported benchmarks updated

  GLM-5.2: 5 benchmark claims (via web search)

• Vendor claim Qwen — reported benchmarks Aug 22, 2026

  Alibaba reported benchmarks updated

  Qwen3.8-Max: 8 benchmark claims (via web search)

• Vendor claim DeepSeek — reported benchmarks Aug 22, 2026

  DeepSeek reported benchmarks updated

  DeepSeek-V4-Pro-Max: 4 benchmark claims (via web search)

• Vendor claim xAI — reported benchmarks Aug 22, 2026

  xAI reported benchmarks updated

  Grok 4.6: 12 benchmark claims (via web search)

• Vendor claim Google — reported benchmarks Aug 22, 2026

  Google reported benchmarks updated

  Gemini 3.1 Pro: 11 benchmark claims (via web search)

• Vendor claim Anthropic — reported benchmarks Aug 22, 2026

  Anthropic reported benchmarks updated

  Claude Opus 5: 12 benchmark claims (via web search)

• Benchmark LMArena (text) Aug 22, 2026

  LMArena (text) scores changed

  Arena Elo (text, overall) updated.

• Model NVIDIA Nemotron Aug 22, 2026

  NVIDIA Nemotron models changed

  NVIDIA-Nemotron-Parse-2.0 license changed from 'other' to 'openmdw-1.1'

• Model Google Gemma Aug 20, 2026

  Google Gemma models changed

  Removed six T5-Gemma model variants with gemma license dated 2025-06-19.

• Benchmark LMArena (text) Aug 20, 2026

  LMArena (text) scores changed

  Arena Elo (text, overall) updated.

• Benchmark GPQA Diamond (Epoch) Aug 20, 2026

  GPQA Diamond (Epoch) scores changed

  GPQA Diamond accuracy (0–1) updated.

• Silent update Epoch Capabilities Index Aug 19, 2026

  deepseek-coder-1.3b-base: Epoch Capabilities Index (ECI) ↑ 62.02 → 62.28

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  Cerebras-GPT-13B: Epoch Capabilities Index (ECI) ↑ 81.66 → 81.79

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  starcoder2-3b: Epoch Capabilities Index (ECI) ↑ 87.33 → 87.44

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  deepseek-coder-6.7b-base: Epoch Capabilities Index (ECI) ↑ 88.2 → 88.31

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  dolly-v2-12b: Epoch Capabilities Index (ECI) ↑ 88.25 → 88.36

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  Baichuan-7B: Epoch Capabilities Index (ECI) ↑ 89.05 → 89.15

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  phi-1_5: Epoch Capabilities Index (ECI) ↑ 90.19 → 90.29

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  xgen-7b-8k-base: Epoch Capabilities Index (ECI) ↑ 92.09 → 92.19

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  starcoder2-7b: Epoch Capabilities Index (ECI) ↑ 92.24 → 92.33

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  gemma-2b: Epoch Capabilities Index (ECI) ↑ 92.91 → 93

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.

• Silent update Epoch Capabilities Index Aug 19, 2026

  mpt-7b: Epoch Capabilities Index (ECI) ↑ 93.31 → 93.4

  Same model id, score moved on Epoch Capabilities Index — a silent re-evaluation or model swap.