November 2026 rankings

In the November 2026 rankings release, Claude Fable 5.2 is #1 Overall at 91.41 (Estimated), out of 214 models ranked. By board, Claude Fable 5.2 is #1 on the Overall, Coding, Reasoning, Agentic, Knowledge and Multimodal boards; Claude Opus 4.8 is #1 on the Math board. The Token Price Index stood at 15 for November 2026 (March 2023 = 100). The release was frozen on November 12, 2026 from commit 04172aa27, so it will not change.

Supported means we have enough independent evidence to stand behind the row. Estimated means the rank is visible and the uncertainty is wider.

Overall

Today's board

Top 10 of 214 models ranked on the Overall board in November 2026.

Overall board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	91.41	Estimated
#2	Gemini 4 Carbon	Google	87.91	Supported
#3	Claude Opus 5.5	Anthropic	85.77	Supported
#4	GPT-6 Astra	OpenAI	84.98	Supported
#5	GPT-6.1 Sol	OpenAI	84.91	Supported
#6	Gemini 4 Argon	Google	81.38	Supported
#7	Claude Sonnet 5.5	Anthropic	81.24	Supported
#8	Claude Fable 5.1	Anthropic	80.82	Supported
#9	Claude Fable 5	Anthropic	80.38	Supported
#10	GPT-6.1 Terra	OpenAI	80.13	Estimated

Coding

Today's board

Top 10 of 158 models ranked on the Coding board in November 2026.

Coding board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	86.31	Estimated
#2	Claude Opus 5.5	Anthropic	82.4	Supported
#3	Claude Sonnet 5.5	Anthropic	81.41	Supported
#4	Gemini 4 Carbon	Google	80.52	Estimated
#5	Claude Fable 5.1	Anthropic	76.34	Supported
#6	GPT-6 Astra	OpenAI	73.91	Supported
#7	Gemini 4 Argon	Google	71.95	Supported
#8	Claude Fable 5	Anthropic	70.97	Supported
#9	Muse Wonder Full	Meta	70.79	Supported
#10	Claude Opus 5	Anthropic	70.3	Supported

Reasoning

Today's board

Top 10 of 158 models ranked on the Reasoning board in November 2026.

Reasoning board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	89.93	Supported
#2	Gemini 4 Carbon	Google	83.37	Supported
#3	GPT-6 Astra	OpenAI	80.43	Supported
#4	Claude Opus 5.5	Anthropic	79.95	Supported
#5	GPT-6.1 Sol	OpenAI	79.94	Supported
#6	Gemini 4 Argon	Google	77.32	Supported
#7	Claude Fable 5.1	Anthropic	76.78	Supported
#8	Claude Sonnet 5.5	Anthropic	76.1	Supported
#9	Muse Wonder Full	Meta	75.88	Supported
#10	GPT-6.1 Terra	OpenAI	75.3	Supported

Agentic

Today's board

Top 10 of 132 models ranked on the Agentic board in November 2026.

Agentic board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	90.06	Estimated
#2	Claude Opus 5.5	Anthropic	89.34	Supported
#3	Claude Fable 5.1	Anthropic	80.35	Supported
#4	Gemini 4 Argon	Google	77.59	Supported
#5	Claude Opus 5	Anthropic	76.98	Supported
#6	GPT-6.1 Terra	OpenAI	76.01	Estimated
#7	GPT-6.1 Sol	OpenAI	75.56	Estimated
#8	Claude Fable 5	Anthropic	74.02	Supported
#9	GPT-6 Astra	OpenAI	73.52	Supported
#10	Kimi K3.1 Pro	Moonshot AI	72.72	Supported

Knowledge

Today's board

Top 10 of 179 models ranked on the Knowledge board in November 2026.

Knowledge board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	97.41	Supported
#2	Claude Opus 5.5	Anthropic	88.17	Supported
#3	Claude Fable 5.1	Anthropic	84.7	Supported
#4	GPT-6.1 Sol	OpenAI	84.13	Supported
#5	GPT-6 Astra	OpenAI	83.81	Supported
#6	Claude Opus 5	Anthropic	80.31	Supported
#7	Claude Fable 5	Anthropic	80.16	Supported
#8	Claude Sonnet 5.5	Anthropic	79.9	Supported
#9	Gemini 4 Argon	Google	79.36	Supported
#10	GPT-6 Sol	OpenAI	78.67	Estimated

MultimodalPreliminary

Today's board

Top 10 of 67 models ranked on the Multimodal board in November 2026. This board is preliminary: its roster is still thin, so it issues no badges yet.

Multimodal board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Fable 5.2	Anthropic	89.35	Estimated
#2	Claude Opus 5.5	Anthropic	84.44	Supported
#3	Claude Fable 5.1	Anthropic	81.2	Estimated
#4	GPT-6 Astra	OpenAI	78.93	Supported
#5	GPT-6.1 Sol	OpenAI	78.36	Supported
#6	Claude Opus 5	Anthropic	77.78	Estimated
#7	Claude Sonnet 5.5	Anthropic	77.65	Estimated
#8	GPT-6.1 Terra	OpenAI	77.22	Supported
#9	GPT-5.6 Sol	OpenAI	75.48	Supported
#10	Claude Fable 5	Anthropic	75.26	Supported

Math

Today's board

Top 10 of 12 models ranked on the Math board in November 2026. This board still uses the legacy weighted category score, so its rows carry no Supported label and issue no badges.

Math board, November 2026: rank, model, lab, score and evidence label for the top 10.
Rank	Model	Lab	Score	Evidence
#1	Claude Opus 4.8	Anthropic	79.3	Estimated
#2	Qwen3.7 Max	Alibaba	78.3	Estimated
#3	GLM-5.2	Z.AI	74.6	Estimated
#4	Kimi K2.6	Moonshot AI	73.2	Estimated
#5	Inkling-Small	Thinking Machines Lab	64.6	Estimated
#6	Qwen3.6 Plus	Alibaba	62.3	Estimated
#7	Qwen3.6-27B	Alibaba	56.2	Estimated
#8	GLM-5.1	Z.AI	55.7	Estimated
#9	Inkling	Thinking Machines Lab	54.6	Estimated
#10	Claude Opus 4.5	Anthropic	52.9	Estimated

Cite this release

BenchLM.ai, "November 2026 rankings", https://benchlm.ai/monthly-rankings/2026-11, frozen November 12, 2026.

@misc{benchlm_release_2026_11,
  author = {{BenchLM.ai}},
  title = {{BenchLM monthly rankings: November 2026}},
  year = {2026},
  month = {11},
  url = {https://benchlm.ai/monthly-rankings/2026-11},
  note = {Frozen 2026-11-12 from commit 04172aa27}
}

Badges

This release issues 42 rank badges, one for each Supported row in the top 10 of a board that issues them. Images and the HTML to copy are on rank badges.

How this release is made

Frozen on November 12, 2026 from commit 04172aa27.
The ranks come from the same method as the live boards, described in the methodology.
Corrections are listed on this page, never edited silently.
The next release comes out early next month.
No payment changes a score, a rank or a label. See our editorial policy.