LLM Reference
activeTool use

MCP-Atlas

Metric: Score (higher is better)

Benchmark for evaluating model performance on Model Context Protocol tool use tasks High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

19

tracked on this benchmark

Score band

84.2 – 29.5

best → lowest tracked

Snapshot trend

+13.60

Jun 18 → Jul 16 · 1 models

Leaderboard

Tracked models ranked by Score (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Kimi K3
Version: 500-task public subset; 100-turn limit; Gemini 3.1 Pro judge; max reasoning effortHarness: Not recordedEvaluator: Not recordedObserved: Jul 16, 2026Confidence: Not recordedSource

Notes: Model variant: Kimi K3 (max reasoning effort). Benchmark variant: MCP-Atlas 500-task public subset with a 100-turn limit. Harness/evaluator: Gemini 3.1 Pro judge; remaining harness/evaluator details not disclosed by Moonshot. Source posture: Moonshot/Kimi self-reported launch table; confidence: medium because the value is vendor-reported and not independently reproduced. Recommended seed value: 84.2.

84.2
2
Gemini 3.5 Flash
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
83.6
3
Claude Opus 4.8
Version: MCP-AtlasHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource

Notes: Low-confidence secondary-source row; useful as supporting Anthropic MCP-native evidence.

82.2
4
Claude Opus 4.7
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
77.3
5
GLM-5.2
Version: MCP-AtlasHarness: Not recordedEvaluator: Not recordedObserved: Jun 13, 2026Confidence: Not recordedSource
76.8
6
Qwen3.7-Max
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
76.4
7
Kimi K2.7-Code
Version: official, Moonshot-reportedHarness: Not recordedEvaluator: Not recordedObserved: Jun 12, 2026Confidence: Not recordedSource

Notes: DAT-6271: Moonshot-run MCP tool-use benchmark. Compare with other MCP Atlas rows cautiously when source harnesses differ.

76.0
8
GPT-5.5
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
75.3
9
GPT-5.5 Pro
Version: MCP-AtlasHarness: Not recordedEvaluator: Not recordedObserved: Apr 24, 2026Confidence: Not recordedSource

Notes: Low-confidence secondary-source row; useful as supporting tool-use evidence.

75.3
10
MiniMax M3
Version: MiniMax-reported public setHarness: Not recordedEvaluator: Not recordedObserved: Jun 1, 2026Confidence: Not recordedSource
74.2
11
Qwen3.6-Plus
Version: llm-stats shows 0 (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
74.1
12
GLM-5.1
Version: independent evaluation (BenchLM, June 2026)Harness: Not recordedEvaluator: Not recordedObserved: Jun 18, 2026Confidence: Not recordedSource
71.8
13
DeepSeek V4 Pro
Version: independent evaluation (BenchLM, June 2026)Harness: Not recordedEvaluator: Not recordedObserved: Jun 18, 2026Confidence: Not recordedSource
69.4
14
Gemini 3.1 Pro Preview
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
69.2
15
DeepSeek V4 Flash
Version: MCPAtlas pass@1, V4-Flash Think MaxHarness: Not recordedEvaluator: Not recordedObserved: Jul 3, 2026Confidence: Not recordedSource
69.0
16
Claude Opus 4.6
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
62.7
17
Claude Sonnet 4.6
Version: llm-stats shows 0 (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
61.3
18
Kimi K2.6
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
55.9
19
Kimi K2.5
Version: MCP-Atlas (accuracy%)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
29.5

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the MCP-Atlas benchmark measure?

Benchmark for evaluating model performance on Model Context Protocol tool use tasks On this page it lists 19 tracked model variants where higher is better.

Is a higher MCP-Atlas score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this MCP-Atlas data?

This benchmark was last reviewed on May 20, 2026. The tracked score average moved +13.60 points across the last 3 snapshots.

Related benchmarks

Last reviewed: May 20, 2026

Resources