BlogLLM Benchmark Scores as Published, Side by Side
SpreadsheetBench: Results as Published
Every result the catalog holds for SpreadsheetBench: 13 models from 7 developers, read from 2 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 Knowledge WorkMaintainer RUCKBReasoning (GitHub owner of the official repository); the arXiv abstract page names the authors, led by Zeyao MaIntroduced 2024Data version 2026103001, as of 2026-10-30
SpreadsheetBench tests whether an AI model can carry out real-world spreadsheet manipulation. The 912 instructions were collected from online Excel forums and cover spreadsheets with multiple tables, non-standard relational tables and non-textual elements. Each instruction has several spreadsheet files as test cases, 2,729 in total (about three per instruction), and outputs are judged in an online-judge style, with formulas recalculated before the check. The dataset is released under CC BY-SA 4.0.
SpreadsheetBench v1
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.5OpenAI, released 2026-04-23 | 88.11% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| Claude Opus 4.7Anthropic, released 2026-04-16 | 88.49% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| Gemini 3.1 Pro PreviewGoogle, released 2026-02-19 | 56.06% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| DeepSeek-V4-Pro (Preview)DeepSeek, released 2026-04-24 | 84.9% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| Kimi K2.6Moonshot AI, released 2026-04-20 | 84.5% | MiniMax, MiniMax M3checked 2026-10-15 | Effort: thinking; Harness: Claude Code |
| GLM-5.1Z.ai, released 2026-04-07 | 85.2% | MiniMax, MiniMax M3checked 2026-10-15 | Effort: thinking; Harness: Claude Code |
| MiniMax M3MiniMax, released 2026-06-01 | 89.35% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
| MiniMax M2.7MiniMax, released 2026-03-18 | 84.92% | MiniMax, MiniMax M3checked 2026-10-15 | Harness: Claude Code |
SpreadsheetBench 2
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.6 SolOpenAI, released 2026-07-09 | 32.4% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Harness: Codex |
| GPT-5.5OpenAI, released 2026-04-23 | 29.1% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: xhigh; Harness: Codex |
| Claude Fable 5Anthropic, released 2026-06-09 | 34.7% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code; Fallback: w/ fallback |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 31.6% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code |
| Kimi K3Moonshot AI, released 2026-07-16 | 34.8% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code |
| GLM-5.2Z.ai, released 2026-06-16 | 28.1% | Moonshot AI, Kimi K3 model cardchecked 2026-10-15 | Effort: max; Harness: Claude Code |
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.