Benchmark Almanac

ToolMaze

Gemini 3.1 Pro 53% leads since June 2026
Not in model cards

Agents complete tool-use tasks while the runtime introduces explicit or hidden tool failures. A judge checks completion across four task structures and five perturbation settings. Metrics report task success, recovery rate, and recovery cost.

Recovery rate includes only runs that encountered a perturbed call. Recovery cost compares actual calls with an oracle recovery path. The runtime includes 273 tool plugins.

Released
June 2026
Kind
Function calling
Level
University
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Items from
Generated
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
April 2026
Scores refreshed
November 19th, 2026

In the authors’ words

Existing benchmarks evaluate Tool-Integrated Reasoning (TIR) in LLMs on idealized ”happy paths”, largely overlooking real-world tool failures. We introduce ToolMaze, a benchmark for dynamic path discovery and error recovery in TIR agents. To separate systematic replanning from blind trial-and-error, ToolMaze adopts a two-dimensional design: DAG-based topological complexity and a 2 × 2 taxonomy of tool perturbations (explicit/implicit, transient/permanent). Evaluations show that perturbations degrade performance across nearly all models, with the sharpest drops under implicit semantic failures. Driven by systemic over-trust in corrupted outputs, Perturbation Recovery Rate (PRR) plummets by around 37% in these scenarios, while complex topologies trap agents in futile trial-and-error loops. Crucially, agentic fault-tolerance improves with model scale 3.66× slower than basic task execution, highlighting dynamic replanning as a distinct bottleneck unaddressed by model scaling or prompting. Data and code are available at https://github.com/Zhudongsheng75/ToolMaze.

Abstract of When Tools Fail: Benchmarking Dynamic Replanning and Anomaly Recovery in LLM Agents (arXiv:2606.05806, CC0)
9
Models scored in its paper

Gemini 3.1 Pro leads with 53% among models released by April 2026

Best score per model from the benchmark's paper, read on November 19th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini 3.1 Pro w/ hint 53%
  2. 2 GPT-5.5 w/ hint 50%
  3. 3 Claude Sonnet 4.6 w/ hint 46.1%
  4. 4 DeepSeek-V4-Pro w/ hint 45.1%
  5. 5 GLM-5.1 w/ hint 38.1%
  6. 6 Qwen3.5 397B-A17B w/ hint 35.6%
  7. 7 Qwen3.6 27B w/ hint 34.1%
  8. 8 Qwen3.5-35B-A3B w/ hint 32.6%
  9. 9 MiniMax-M2.7 w/ hint 27.5%

Source: The ToolMaze paper

#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 53% w/ hint Feb 2026 Official leaderboard
2 GPT-5.5 50% w/ hint Apr 2026 Official leaderboard
3 Claude Sonnet 4.6 46.1% w/ hint Feb 2026 Official leaderboard
4 DeepSeek-V4-Pro 45.1% w/ hint Apr 2026 Official leaderboard
5 GLM-5.1 38.1% w/ hint Apr 2026 Official leaderboard
6 Qwen3.5 397B-A17B 35.6% w/ hint Feb 2026 Official leaderboard
7 Qwen3.6 27B 34.1% w/ hint Apr 2026 Official leaderboard
8 Qwen3.5-35B-A3B 32.6% w/ hint Feb 2026 Official leaderboard
9 MiniMax-M2.7 27.5% w/ hint Mar 2026 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.

Lowest effort to highest, light to dark
Gemini 3.1 Pro
w/o hint w/ hint
53%
GPT-5.5
w/o hint w/ hint
50%
Claude Sonnet 4.6
w/o hint w/ hint
46.1%
DeepSeek-V4-Pro
w/o hint w/ hint
45.1%
GLM-5.1
w/o hint w/ hint
38.1%
Qwen3.5 397B-A17B
w/o hint w/ hint
35.6%
Qwen3.6 27B
w/o hint w/ hint
34.1%
Qwen3.5-35B-A3B
w/o hint
32.6%
MiniMax-M2.7
w/o hint
27.5%

Source: The ToolMaze paper

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro w/ hint 53%
GPT-5.5 w/ hint 50%
Claude Sonnet 4.6 w/ hint 46.1%
DeepSeek-V4-Pro w/ hint 45.1%
GLM-5.1 w/ hint 38.1%
DeepSeek-V4-Pro w/o hint 36.3%
GPT-5.5 w/o hint 35.9%
Qwen3.5 397B-A17B w/ hint 35.6%
Qwen3.6 27B w/ hint 34.1%
Qwen3.5-35B-A3B w/ hint 32.6%
GLM-5.1 w/o hint 32.3%
Gemini 3.1 Pro w/o hint 32.2%
Qwen3.5 397B-A17B w/o hint 31.6%
Qwen3.5-35B-A3B w/o hint 31.6%
Qwen3.6 27B w/o hint 29.6%
MiniMax-M2.7 w/ hint 27.5%
Claude Sonnet 4.6 w/o hint 27.4%
MiniMax-M2.7 w/o hint 25%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Google DeepMind
Benchmark paper Not tested since Jun 2026

Source: The ToolMaze paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemini 3.1 Pro Claude Sonnet 4.6 Qwen3.5 397B-A17B 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemini 3.1 Pro Claude Sonnet 4.6

Source: Official ToolMaze leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Tool use

All 265 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Neon 56.3%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (9 entries).
  • Pass Models scored 9 models we track.
  • Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 53%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 18 rows for models we track at the last read.
  • Pass Last read November 19th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check