Meeseeks Utility Rate
Models follow complex instructions with feedback for up to 20 turns. The score measures fully usable responses as a percentage.
The paper reports utility rate separately from requirement-level Meeseeks Score. The importer reads final-turn utility rate.
- Made by
- Meituan
- Released
- April 2025
- Kind
- Multi-turn conversation
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Strongest model tested
- Claude Opus 4 #108 on the capability index
- Last model added
- August 2025
- Scores refreshed
- November 12th, 2026
Claude Sonnet 4 leads with 90.5% among models released by August 2025
Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
The newest model with a score here came out in August 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 Claude Sonnet 4 20-turn feedback budget, thinking 90.5%
- 2 Claude Opus 4 20-turn feedback budget 87.6%
- 3
Qwen3-235B-A22B 20-turn feedback budget, thinking 87.1% - 4 GLM-4.5 20-turn feedback budget 86.1%
- 5 o3-mini 20-turn feedback budget, high 84.6%
- 6
DeepSeek-V3.1 20-turn feedback budget 79.1% - 7
Doubao-Seed-1.6 20-turn feedback budget 76.9% - 8 GPT-4.1 20-turn feedback budget 69.4%
- 9 Gemini 2.5 Flash (snapshot unspecified) 20-turn feedback budget 67.2%
- 10 Qwen3-32B 20-turn feedback budget 66.4%
- 11 Kimi-K2-Instruct 20-turn feedback budget 62.4%
- 12 Qwen3-14B 20-turn feedback budget, thinking 58%
Source: Official Meeseeks Utility Rate leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4 | 90.5% | 20-turn feedback budget, thinking | May 2025 | Official leaderboard |
| 2 | Claude Opus 4 | 87.6% | 20-turn feedback budget | May 2025 | Official leaderboard |
| 3 | Qwen3-235B-A22B | 87.1% | 20-turn feedback budget, thinking | Apr 2025 | Official leaderboard |
| 4 | GLM-4.5 | 86.1% | 20-turn feedback budget | Aug 2025 | Official leaderboard |
| 5 | o3-mini | 84.6% | 20-turn feedback budget, high | Jan 2025 | Official leaderboard |
| 6 | DeepSeek-V3.1 | 79.1% | 20-turn feedback budget | Aug 2025 | Official leaderboard |
| 7 | Doubao-Seed-1.6 | 76.9% | 20-turn feedback budget | Jun 2025 | Official leaderboard |
| 8 | GPT-4.1 | 69.4% | 20-turn feedback budget | Apr 2025 | Official leaderboard |
| 9 | Gemini 2.5 Flash (snapshot unspecified) | 67.2% | 20-turn feedback budget | Official leaderboard | |
| 10 | Qwen3-32B | 66.4% | 20-turn feedback budget | Apr 2025 | Official leaderboard |
| 11 | Kimi-K2-Instruct | 62.4% | 20-turn feedback budget | Official leaderboard | |
| 12 | Qwen3-14B | 58% | 20-turn feedback budget, thinking | Apr 2025 | Official leaderboard |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Qwen3-235B-A22B Qwen3-32B Qwen3-14B Source: Official Meeseeks Utility Rate leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Sonnet 4 | 20-turn feedback budget, thinking | 90.5% | |||
| Claude Opus 4 | 20-turn feedback budget | 87.6% | |||
| Claude Opus 4 | 20-turn feedback budget, thinking | 87.6% | |||
| Qwen3-235B-A22B | 20-turn feedback budget, thinking | 87.1% | |||
| GLM-4.5 | 20-turn feedback budget | 86.1% | |||
| o3-mini | 20-turn feedback budget, high | 84.6% | |||
| Claude Sonnet 4 | 20-turn feedback budget | 79.4% | |||
| DeepSeek-V3.1 | 20-turn feedback budget | 79.1% | |||
| Doubao-Seed-1.6 | 20-turn feedback budget | 76.9% | |||
| Qwen3-235B-A22B | 20-turn feedback budget | 73.9% | |||
| GPT-4.1 | 20-turn feedback budget | 69.4% | |||
| Qwen3-32B | 20-turn feedback budget | 66.4% | |||
| Qwen3-14B | 20-turn feedback budget, thinking | 58% | |||
| Qwen3-32B | 20-turn feedback budget, thinking | 57.2% | |||
| Qwen3-14B | 20-turn feedback budget | 47% |
3 models have held number one on the official leaderboard since its release in April 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Meeseeks Utility Rate leaderboard
Qwen3-235B-A22B on the official leaderboard passed 85% in April 2025
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Meeseeks Utility Rate leaderboard
No lab model card reports it yet
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.
More in Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
- IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose).
Nemotron-Cascade 2 30B-A3B 82.9% - MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85% - IFStruct In use Models generate JSON or YAML that must follow a requested schema and phrasing constraints, without constrained decoding. A response passes only if every structural constraint holds. Gemma 4 31B IT 95.9%