BlogLLM Benchmark Scores as Published, Side by Side
ArXivMath: Results as Published
Every result the catalog holds for ArXivMath: 8 models from 3 developers, read from 3 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Science and MathMaintainer MathArenaData version 2026103001, as of 2026-10-30
Final-answer research-level mathematics problems extracted monthly from recent arXiv papers, filtered for being self-contained, non-trivial and verifiable.
Not checked against a maintainer's page.
ArXivMath (August 2026), with tools
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-6 AstraOpenAI, released 2026-09-03 | 88.6% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25 | Tools: enabled |
| Claude Sonnet 5.5Anthropic, released 2026-09-28 | 95.2% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: max; Tools: Code execution sandbox, no internet access |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 96.9% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25; printed on 1 more page | Effort: max; Tools: enabled |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 92.1% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25; printed on 1 more page | Effort: max; Tools: enabled |
| 87.7% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25 | Effort: high; Tools: enabled | |
| Claude Opus 5Anthropic, released 2026-07-24 | 90.4% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25 | Effort: max; Tools: enabled |
ArXivMath (August 2026), no tools
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Sonnet 5.5Anthropic, released 2026-09-28 | 86.8% | Anthropic, Claude Sonnet 5.5 System Cardchecked 2026-10-19 | Effort: max |
| Claude Opus 5.5Anthropic, released 2026-09-22 | 91.2% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25; printed on 1 more page | Effort: max; Tools: none |
| Claude Fable 5.1Anthropic, released 2026-09-01 | 82.9% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25; printed on 1 more page | Effort: max; Tools: none |
| Claude Opus 5Anthropic, released 2026-07-24 | 78.1% | Anthropic, Claude Opus 5.5 System Cardchecked 2026-10-25 | Effort: max; Tools: none |
ArXivMath (June 2026), without tools
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.6 SolOpenAI, released 2026-07-09 | 86.73% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Tools: none |
| Claude Mythos 5.1Anthropic, released 2026-09-01 | 91.33% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Effort: max; Tools: none |
| Gemini 3.1 Pro PreviewGoogle, released 2026-02-19 | 65.99% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Tools: none |
ArXivMath (June 2026), with tools
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Mythos 5.1Anthropic, released 2026-09-01 | 93.88% | Anthropic, System Card: Claude Fable 5.1 & Claude Mythos 5.1checked 2026-10-25 | Effort: max; Tools: enabled |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.