Task ranking
Best LLM for Math & Reasoning
Claude Opus 5.5 ranks first for Math & Reasoning as of November 15, 2026.[1][2][3][4][5][6][7] Leads on LiveBench Mathematics with 97.08% and ties for first on OTIS Mock AIME 2024-2025 with 100%. Gemini 4 Carbon scores within two points, so read the top two as close rather than as a measured gap. Gemini 4 Carbon is the next-ranked alternative. Ranks first on LMArena overall text arena and instruction-following, leads on GPQA Diamond with 96.5%, and ties for first on OTIS Mock AIME with 100%.
About this recommendation
- Updated
- Evidence through
- Sources
- 20
- Revision
- v133
Decision audit
Why this result
Inspect the inputs and the computed order behind the recommendation.
Models screened
24
live candidates
Evaluation feeds
7
task-weighted
Winner coverage
100%
intended feed weight
Largest provider share
3 of 8
Anthropic: Claude Opus 5.5 leads the measured task comparison with 3 independent evaluation families. Community volume, release age and usage popularity add no ranking points. The leading scores are close under this policy.
Real-world track record is still developing. The recommendation can lead on independent measurements while practitioner reports accumulate.
Sources evaluated
These are the weights used in scoring. Missing measurements are omitted from each model’s average. A feed with no result for any model released in the last 60 days is stale and does not score until it catches up.
| Evaluation feed | Weight | Winner result | Field measured |
|---|---|---|---|
| LiveBench Mathematics (2026-06-25)Observed 2026-10-18 · evaluation date not supplied · claude-opus-5-5-max-effortfailing; last successful check 2026-10-18 | 28% | #1 | 17/24 |
| LiveBench Reasoning (2026-06-25)Observed 2026-10-18 · evaluation date not supplied · claude-opus-5-5-max-effortfailing; last successful check 2026-10-18 | 20% | #3 | 17/24 |
| LMArena MathObserved 2026-11-11 · evaluation date not supplied · high effort | 20% | #5 | 22/24 |
| GPQA Diamond (Epoch AI)Observed 2026-09-22 · evaluated 2026-09-22 · max effort · standard error 1.90 | 12% | #34 | 22/24 |
| FrontierMath Tiers 1-3 (Epoch AI)Observed 2026-09-22 · evaluated 2026-09-22 · max effort · standard error 1.70 | 8% | #4 | 20/24 |
| LMArena Instruction FollowingObserved 2026-11-11 · evaluation date not supplied · high effort | 8% | #3 | 23/24 |
| OTIS Mock AIME (Epoch AI)Observed 2026-09-22 · evaluated 2026-09-22 · max effort · standard error 0.00 | 4% | #1 | 23/24 |
Provider concentration
Each exact model is scored separately; provider identity is not a ranking input.
- Google3 models
- Anthropic2 models
- OpenAI2 models
- Meta1 model
Decision table
Every published model is shown in computed order. Practitioner sources are distinct community threads, not the citations repeated in the prose below.
Anthropic · coverage 100%
#1 OTIS Mock AIME (Epoch AI) · #1 LiveBench Mathematics
Google · coverage 32% · close score
#1 GPQA Diamond (Epoch AI) · #1 OTIS Mock AIME (Epoch AI)
Google · coverage 52% · close score
#1 OTIS Mock AIME (Epoch AI) · #1 LMArena Math
OpenAI · coverage 100%
#1 OTIS Mock AIME (Epoch AI) · #2 FrontierMath Tiers 1-3 (Epoch AI)
OpenAI · coverage 100%
#1 OTIS Mock AIME (Epoch AI) · #1 LiveBench Reasoning
Anthropic · coverage 100%
#1 OTIS Mock AIME (Epoch AI) · #4 GPQA Diamond (Epoch AI)
Meta · coverage 88%
#9 LiveBench Mathematics · #12 LiveBench Reasoning
Google · coverage 100%
#7 GPQA Diamond (Epoch AI) · #9 LMArena Math
Task score averages the available, task-weighted measurements. Benchmark positions compare models matched to our catalog. Models measured by at least two independent evaluator families are considered first. Missing tests, practitioner report counts, popularity and release age add no points. Coverage describes evidence breadth, not a probability of correctness. Scores within two points are shown as close under this policy; this is not a statistical confidence interval. When the leading scores are that close, the model measured by more evaluator families ranks first.
- 01
Claude Opus 5.5Anthropic · $4 in / $20 out per M tokens · 1M context
Leads on LiveBench Mathematics with 97.08% and ties for first on OTIS Mock AIME 2024-2025 with 100%.
Best when: Use for competition mathematics: scores 100% on OTIS Mock AIME 2024-2025 and 97.08% on LiveBench Mathematics (#1 of 59).More tips
- Use for research-level mathematics: solves 91.2% of FrontierMath tiers 1-3 (#4 of 76).
No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 02
Gemini 4 CarbonGoogle · $2 in / $10 out per M tokens · 1.0M context
Ranks first on LMArena overall text arena and instruction-following, leads on GPQA Diamond with 96.5%, and ties for first on OTIS Mock AIME with 100%.
Best when: Use for graduate-level science questions: leads with 96.5% on GPQA Diamond (#1 of 113).More tips
- Use for competition mathematics: ties for first with 100% on OTIS Mock AIME 2024-2025.
No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 03
Gemini 4 ArgonGoogle · $2 in / $10 out per M tokens · 1.0M context
Leads LMArena maths category with Elo 1544, ties for first on OTIS Mock AIME with 100%, and scores 96.4% on GPQA Diamond (#2 of 113).
Best when: Use for math-heavy conversational tasks: ranks #1 on LMArena maths category with Elo 1544.More tips
- Use for graduate-level science: scores 96.4% on GPQA Diamond (#2 of 113).
No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 04
GPT-6.1 SolOpenAI · $2 in / $10 out per M tokens · 1.1M context
GPT-6.1 Sol scores 100% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#1 of 114), competition mathematics.
Best when: Its strongest result for this task is on OTIS Mock AIME, which measures competition mathematics. - 05
GPT-6 AstraOpenAI · $10 in / $50 out per M tokens · 1.1M context
Leads LiveBench Reasoning with 92.65%, ties for second on FrontierMath with 93.7%.
Best when: Use for multi-step reasoning evaluation: ranks #1 on LiveBench Reasoning with 92.65%.More tips
- Use for research-level mathematics: solves 93.7% of FrontierMath tiers 1-3 (#2 of 76).
Watch out for
- Watch for research paper reproduction: PaperBenchX evaluation reported only 13.98% paper complete reproduction rate.
- 06
Claude Sonnet 5.5Anthropic · $2 in / $10 out per M tokens · 1M context
Ties for first on OTIS Mock AIME with 100%, scores 95.6% on GPQA Diamond (#4 of 113), and ranks sixth on LiveBench Reasoning with 91.63%.
Best when: Use for competition mathematics: ties for first with 100% on OTIS Mock AIME 2024-2025.More tips
- Use for graduate-level science: scores 95.6% on GPQA Diamond (#4 of 113).
No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 07
Muse Spark 1.3Meta · $1.25 in / $4.25 out per M tokens · 1.0M context
Muse Spark 1.3 scores 95.95% on LiveBench Mathematics (#9 of 59), using objectively scored competition and olympiad-style tasks.
Best when: Its strongest result for this task is on LiveBench Mathematics, which measures mathematical problem solving.No practitioner reports are linked yet, so no cautions are listed. Absence of reports is not evidence of reliability.
- 08
Gemini 3.7 FlashGoogle · $0.75 in / $3.75 out per M tokens · 1.0M context
Ranks ninth on LMArena maths category with Elo 1506 and scores 94.8% on GPQA Diamond (#7 of 113), showing strong performance for a mid-tier model.
Best when: Use for graduate-level science: scores 94.8% on GPQA Diamond (#7 of 113).More tips
- Use for competition mathematics: scores 97.2% on OTIS Mock AIME 2024-2025 (#30 of 114).
Watch out for
- Watch for research-level mathematics: solves only 71.6% of FrontierMath tiers 1-3 (#23 of 76).
Frequently asked
Sources
- 1
“Within the 59 catalog models matched to this evaluation: Scores 97.08% on LiveBench Mathematics (#1 of 59), using objectively scored competition and olympiad-style tasks.”
LiveBench Mathematics · Benchmark · Oct 18, 2026 - 2
“Within the 114 catalog models matched to this evaluation: Scores 100% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#1 of 114), competition mathematics.”
OTIS Mock AIME (Epoch AI) · Benchmark · Sep 22, 2026 - 3
“Within the 59 catalog models matched to this evaluation: Scores 92.15% on LiveBench Reasoning (#3 of 59), an objective, ground-truth-scored evaluation refreshed with new questions.”
LiveBench Reasoning · Benchmark · Oct 18, 2026 - 4
“Within the 147 catalog models matched to this evaluation: Ranks #1 of 147 on LMArena's overall text arena (Elo 1526), based on blind human preference votes.”
LMArena text arena · Benchmark · Nov 11, 2026 - 5
“Within the 148 catalog models matched to this evaluation: Ranks #1 of 148 on LMArena's instruction-following category (Elo 1539), based on blind human preference votes.”
LMArena instruction-following category · Benchmark · Nov 11, 2026 - 6
“Within the 113 catalog models matched to this evaluation: Scores 96.5% on GPQA Diamond in Epoch AI's independent run (#1 of 113), graduate-level science questions.”
GPQA Diamond (Epoch AI) · Benchmark · Nov 2, 2026 - 7
“Within the 114 catalog models matched to this evaluation: Scores 100% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#1 of 114), competition mathematics.”
OTIS Mock AIME (Epoch AI) · Benchmark · Nov 2, 2026 - 8
“Within the 76 catalog models matched to this evaluation: Solves 91.2% of FrontierMath tiers 1 to 3 in Epoch AI's independent run (#4 of 76), research-level mathematics.”
FrontierMath Tiers 1-3 (Epoch AI) · Benchmark · Sep 22, 2026 - 9
“Within the 143 catalog models matched to this evaluation: Ranks #1 of 143 on LMArena's maths category (Elo 1544), based on blind human preference votes for maths prompts.”
LMArena maths category · Benchmark · Nov 11, 2026 - 10
“Within the 113 catalog models matched to this evaluation: Scores 96.4% on GPQA Diamond in Epoch AI's independent run (#2 of 113), graduate-level science questions.”
GPQA Diamond (Epoch AI) · Benchmark · Oct 13, 2026 - 11
“Within the 114 catalog models matched to this evaluation: Scores 100% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#1 of 114), competition mathematics.”
OTIS Mock AIME (Epoch AI) · Benchmark · Sep 29, 2026 - 12
“Within the 59 catalog models matched to this evaluation: Scores 92.65% on LiveBench Reasoning (#1 of 59), an objective, ground-truth-scored evaluation refreshed with new questions.”
LiveBench Reasoning · Benchmark · Oct 18, 2026 - 13
“Within the 76 catalog models matched to this evaluation: Solves 93.7% of FrontierMath tiers 1 to 3 in Epoch AI's independent run (#2 of 76), research-level mathematics.”
FrontierMath Tiers 1-3 (Epoch AI) · Benchmark · Aug 30, 2026 - 14
AI 新闻月刊(2026年10月)
Original-language excerpt:“2026-10-09 陶哲轩牵头成立 AHM 抵制 OpenAI:700 份机器证明引爆「数学末日」 - 菲尔兹奖得主陶哲轩牵头「人类数学协会」(AHM)发布联合声明,全面抵制 OpenAI,敦促全球数学家停止与其合作。 - 导火索:OpenAI 用内部模型测试约 8000 个开放性数学问题(命中率约 5%、平均每题仅消耗 3 小时 GPT-Pro 级算力),一次性发布 700 多份机器生成证明手稿。 - 手稿包含「唯一博弈猜想」等学者毕生课题的「证明」,被理论计算机科学家 Dana Moshkovitz 评价逻辑混乱;Scott Aaronson 称这场冲击为 Mathocalypse(数学末日)。 - 此前 PaperBenchX 测评指出 GPT-6 Astra 论文完整复现率仅 13.98%;本月另有 25 位菲尔兹奖得主联名呼吁商业公司放慢脚步。 - 相关链接:量子位: 陶哲轩带头宣战!人类数学家联合抵制 OpenAI - 相关链接:Bloom…”
ikaijua · GitHub · Nov 2, 2026 - 15
“Within the 114 catalog models matched to this evaluation: Scores 100% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#1 of 114), competition mathematics.”
OTIS Mock AIME (Epoch AI) · Benchmark · Sep 29, 2026 - 16
“Within the 113 catalog models matched to this evaluation: Scores 95.6% on GPQA Diamond in Epoch AI's independent run (#4 of 113), graduate-level science questions.”
GPQA Diamond (Epoch AI) · Benchmark · Sep 29, 2026 - 17
“Within the 59 catalog models matched to this evaluation: Scores 95.95% on LiveBench Mathematics (#9 of 59), using objectively scored competition and olympiad-style tasks.”
LiveBench Mathematics · Benchmark · Oct 18, 2026 - 18
“Within the 113 catalog models matched to this evaluation: Scores 94.8% on GPQA Diamond in Epoch AI's independent run (#7 of 113), graduate-level science questions.”
GPQA Diamond (Epoch AI) · Benchmark · Aug 14, 2026 - 19
“Within the 114 catalog models matched to this evaluation: Scores 97.2% on the OTIS Mock AIME 2024-2025 set in Epoch AI's independent run (#30 of 114), competition mathematics.”
OTIS Mock AIME (Epoch AI) · Benchmark · Aug 14, 2026 - 20
“Within the 76 catalog models matched to this evaluation: Solves 71.6% of FrontierMath tiers 1 to 3 in Epoch AI's independent run (#23 of 76), research-level mathematics.”
FrontierMath Tiers 1-3 (Epoch AI) · Benchmark · Aug 14, 2026
Rankings synthesized from community evidence and open benchmarks. See methodology. Not driven by vendor marketing.