BlogLLM Benchmark Scores as Published, Side by Side
LOCA-Bench: Results as Published
Every result the catalog holds for LOCA-Bench: 8 models from 5 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 Long ContextMaintainer LOCA-Bench authorsIntroduced 2026Data version 2026103001, as of 2026-10-30
LOCA-Bench tests an agent whose environment description runs to hundreds of thousands of tokens: the agent has to find and use the relevant parts of a long environment specification while acting in ReAct mode. The score is the task pass rate at the stated description length.
Not checked against a maintainer's page.
LOCA-Bench (256k environment description)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPT-5.6 SolOpenAI, released 2026-07-09 | 60% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| Claude Opus 5Anthropic, released 2026-07-24 | 53.7% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| Claude Opus 4.7Anthropic, released 2026-04-16 | 57% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| Kimi K3Moonshot AI, released 2026-07-16 | 49.7% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| GLM-5.3Z.ai, released 2026-08-18 | 50.3% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| MiniMax-M3.1-FlashMiniMax, released 2026-10-19 | 60.3% | MiniMax, MiniMax-M3.1-Flash model cardchecked 2026-10-25 | - |
| MiniMax M3MiniMax, released 2026-06-01 | 49.3% | MiniMax, MiniMax M3checked 2026-10-15 | - |
| MiniMax M2.7MiniMax, released 2026-03-18 | 0% | MiniMax, MiniMax M3checked 2026-10-15 | - |
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.