Toloka Arena Enterprise Tool Use (November 2026)
Agents complete private workflows across 11 industries. They must leave the right records and follow the required process in all five attempts.
Toloka compared ten models across editions and confirmed that the tasks changed. Fireworks publishes a separate comparison of the updated benchmark.
- Made by
- Toloka
- Released
- November 2026
- Kind
- Customer service
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Kept private
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- APIs and tools
- Refresh
- New versions replace it
- Run by
- A host
- Based on
- Toloka Arena
- In views
- Agents
- Strongest model tested
- GPT-6.2 Astra #1 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
Gemini 4 Neon leads with 81% Gemini 4 Neon leads with 81%
Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.
- 81.0
Gemini 4 Neon
- 77.0 Claude Fable 5.2
- 76.0
Gemini 4 Hydrogen
- 75.0 Claude Opus 5.5
- 74.0
Gemini 4 Carbon
- 73.0
Gemini 4 Argon
- 72.0 GPT-6.2 Astra
- 71.0 Claude Fable 5.1
- 67.0 GPT-6 Astra
- 66.0 GPT-6.1 Sol
- 65.7
Muse Wonder
- 65.0
Qwen4-Max
- 65.0 GPT-6.2 Sol
- 64.0 Claude Sonnet 5.5
- 62.0 Claude Opus 5
- 61.0 GPT-5.6 Sol
- 59.0
Gemini 3.8 Flash
- 58.0
Gemini 3.7 Flash
- 58.0 GPT-6 Sol
- 56.0 Grok 4.6
- 55.0 GPT-6.1 Terra
- 52.0
Muse Ember
- 51.0 MiMo-V2.6 Pro
- 49.0 Kimi K3.1 Pro
- 44.0
Qwen3.8 Max (0902)
- 40.0 Reflection Beam
- 36.0
Qwen 3.8 27B
- 32.0
Tencent Hy4
- 31.0 Grok 4.7
- 30.0 GPT-6 Luna
- 29.0
Mistral Large 4
- 26.0
DeepSeek V4.1 Flash
- 22.0
MiniMax-M3.1-Flash
- 21.0 Claude Haiku 5.5
- 19.0 Claude Sonnet 5
- 18.0
Muse Spark Open
- 12.0
Inkling 2 Small
- 9.00
Step 5 Preview
- 8.00
Gemini 3.1 Pro
- 7.00 Kimi K3
- 7.00 GLM-5.3
- 6.00
Nemotron 3.5 Super VL
- 1.00 Flint v1.0
- 0.0 Mellum2.1 12B A2.5B Thinking
- 0.0 Granite 4.2 8B
- 0.0 LFM2.5-24B-A2B
- 1
Gemini 4 Neon 81%
- 2 Claude Fable 5.2 77%
- 3
Gemini 4 Hydrogen 76%
- 4 Claude Opus 5.5 75%
- 5
Gemini 4 Carbon 74%
- 6
Gemini 4 Argon 73%
- 7 GPT-6.2 Astra 72%
- 8 Claude Fable 5.1 71%
- 9 GPT-6 Astra 67%
- 10 GPT-6.1 Sol 66%
- 11
Muse Wonder 65.7%
- 12
Qwen4-Max 65%
- 13 GPT-6.2 Sol 65%
- 14 Claude Sonnet 5.5 64%
- 15 Claude Opus 5 62%
- 16 GPT-5.6 Sol 61%
- 17
Gemini 3.8 Flash 59%
- 18
Gemini 3.7 Flash 58%
- 19 GPT-6 Sol 58%
- 20 Grok 4.6 56%
- 21 GPT-6.1 Terra 55%
- 22
Muse Ember 52%
- 23 MiMo-V2.6 Pro 51%
- 24 Kimi K3.1 Pro 49%
- 25
Qwen3.8 Max (0902) 44%
- 26 Reflection Beam 40%
- 27
Qwen 3.8 27B 36%
- 28
Tencent Hy4 32%
- 29 Grok 4.7 31%
- 30 GPT-6 Luna 30%
- 31
Mistral Large 4 29%
- 32
DeepSeek V4.1 Flash 26%
- 33
MiniMax-M3.1-Flash 22%
- 34 Claude Haiku 5.5 21%
- 35 Claude Sonnet 5 19%
- 36
Muse Spark Open 18%
- 37
Inkling 2 Small 12%
- 38
Step 5 Preview 9.00%
- 39
Gemini 3.1 Pro 8.00%
- 40 Kimi K3 7.00%
- 41 GLM-5.3 7.00%
- 42
Nemotron 3.5 Super VL 6.00%
- 43 Flint v1.0 1.00%
- 44 Mellum2.1 12B A2.5B Thinking 0%
- 45 Granite 4.2 8B 0%
- 46 LFM2.5-24B-A2B 0%
Source: Official Toloka Arena Enterprise Tool Use (November 2026) leaderboard
- 81.0
Gemini 4 Neon
- 77.0 Claude Fable 5.2
- 76.0
Gemini 4 Hydrogen
- 75.0 Claude Opus 5.5
- 74.0
Gemini 4 Carbon
- 73.0
Gemini 4 Argon
- 72.0 GPT-6.2 Astra
- 71.0 Claude Fable 5.1
- 67.0 GPT-6 Astra
- 66.0 GPT-6.1 Sol
- 65.7
Muse Wonder
- 65.0
Qwen4-Max
- 65.0 GPT-6.2 Sol
- 64.0 Claude Sonnet 5.5
- 62.0 Claude Opus 5
- 61.0 GPT-5.6 Sol
- 59.0
Gemini 3.8 Flash
- 58.0
Gemini 3.7 Flash
- 58.0 GPT-6 Sol
- 56.0 Grok 4.6
- 55.0 GPT-6.1 Terra
- 52.0
Muse Ember
- 51.0 MiMo-V2.6 Pro
- 49.0 Kimi K3.1 Pro
- 44.0
Qwen3.8 Max (0902)
- 36.0
Qwen 3.8 27B
- 32.0
Tencent Hy4
- 31.0 Grok 4.7
- 30.0 GPT-6 Luna
- 29.0
Mistral Large 4
- 26.0
DeepSeek V4.1 Flash
- 22.0
MiniMax-M3.1-Flash
- 21.0 Claude Haiku 5.5
- 19.0 Claude Sonnet 5
- 18.0
Muse Spark Open
- 12.0
Inkling 2 Small
- 9.00
Step 5 Preview
- 8.00
Gemini 3.1 Pro
- 7.00 Kimi K3
- 7.00 GLM-5.3
- 6.00
Nemotron 3.5 Super VL
- 0.0 Granite 4.2 8B
- 0.0 LFM2.5-24B-A2B
- 1
Gemini 4 Neon 81%
- 2 Claude Fable 5.2 77%
- 3
Gemini 4 Hydrogen 76%
- 4 Claude Opus 5.5 75%
- 5
Gemini 4 Carbon 74%
- 6
Gemini 4 Argon 73%
- 7 GPT-6.2 Astra 72%
- 8 Claude Fable 5.1 71%
- 9 GPT-6 Astra 67%
- 10 GPT-6.1 Sol 66%
- 11
Muse Wonder 65.7%
- 12
Qwen4-Max 65%
- 13 GPT-6.2 Sol 65%
- 14 Claude Sonnet 5.5 64%
- 15 Claude Opus 5 62%
- 16 GPT-5.6 Sol 61%
- 17
Gemini 3.8 Flash 59%
- 18
Gemini 3.7 Flash 58%
- 19 GPT-6 Sol 58%
- 20 Grok 4.6 56%
- 21 GPT-6.1 Terra 55%
- 22
Muse Ember 52%
- 23 MiMo-V2.6 Pro 51%
- 24 Kimi K3.1 Pro 49%
- 25
Qwen3.8 Max (0902) 44%
- 26
Qwen 3.8 27B 36%
- 27
Tencent Hy4 32%
- 28 Grok 4.7 31%
- 29 GPT-6 Luna 30%
- 30
Mistral Large 4 29%
- 31
DeepSeek V4.1 Flash 26%
- 32
MiniMax-M3.1-Flash 22%
- 33 Claude Haiku 5.5 21%
- 34 Claude Sonnet 5 19%
- 35
Muse Spark Open 18%
- 36
Inkling 2 Small 12%
- 37
Step 5 Preview 9.00%
- 38
Gemini 3.1 Pro 8.00%
- 39 Kimi K3 7.00%
- 40 GLM-5.3 7.00%
- 41
Nemotron 3.5 Super VL 6.00%
- 42 Granite 4.2 8B 0%
- 43 LFM2.5-24B-A2B 0%
Source: Official Toloka Arena Enterprise Tool Use (November 2026) leaderboard
| # | Entry | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 81% ±7.7 | tolokaforge | Nov 2026 | Official leaderboard |
| 2 | Claude Fable 5.2 | 77% ±8.2 | tolokaforge | Nov 2026 | Official leaderboard |
| 3 | | 76% ±8.4 | tolokaforge | Nov 2026 | Official leaderboard |
| 4 | Claude Opus 5.5 | 75% ±8.5 | tolokaforge | Sep 2026 | Official leaderboard |
| 5 | | 74% ±8.6 | tolokaforge | Oct 2026 | Official leaderboard |
| 6 | | 73% ±8.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 7 | GPT-6.2 Astra | 72% ±8.8 | tolokaforge | Nov 2026 | Official leaderboard |
| 8 | Claude Fable 5.1 | 71% ±8.9 | tolokaforge | Sep 2026 | Official leaderboard |
| 9 | GPT-6 Astra | 67% ±9.2 | tolokaforge | Sep 2026 | Official leaderboard |
| 10 | GPT-6.1 Sol | 66% ±9.3 | tolokaforge | Sep 2026 | Official leaderboard |
| 11 | | 65.7% ±9.4 | tolokaforge | Oct 2026 | Official leaderboard |
| 12 | | 65% ±9.3 | tolokaforge | Nov 2026 | Official leaderboard |
| 13 | GPT-6.2 Sol | 65% ±9.3 | tolokaforge | Nov 2026 | Official leaderboard |
| 14 | Claude Sonnet 5.5 | 64% ±9.4 | tolokaforge | Sep 2026 | Official leaderboard |
| 15 | Claude Opus 5 | 62% ±9.5 | tolokaforge | Jul 2026 | Official leaderboard |
| 16 | GPT-5.6 Sol | 61% ±9.6 | tolokaforge | Jul 2026 | Official leaderboard |
| 17 | | 59% ±9.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 18 | | 58% ±9.7 | tolokaforge | Aug 2026 | Official leaderboard |
| 19 | GPT-6 Sol | 58% ±9.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 20 | Grok 4.6 | 56% ±9.7 | tolokaforge | Aug 2026 | Official leaderboard |
| 21 | GPT-6.1 Terra | 55% ±9.8 | tolokaforge | Oct 2026 | Official leaderboard |
| 22 | | 52% ±9.8 | tolokaforge | Nov 2026 | Official leaderboard |
| 23 | MiMo-V2.6 Pro | 51% ±9.9 | tolokaforge | Sep 2026 | Official leaderboard |
| 24 | Kimi K3.1 Pro | 49% ±9.8 | tolokaforge | Oct 2026 | Official leaderboard |
| 25 | | 44% ±9.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 26 | Reflection Beam | 40% ±9.6 | tolokaforge | Official leaderboard | |
| 27 | | 36% ±9.4 | tolokaforge | Aug 2026 | Official leaderboard |
| 28 | | 32% ±9.2 | tolokaforge | Nov 2026 | Official leaderboard |
| 29 | Grok 4.7 | 31% ±9.1 | tolokaforge | Sep 2026 | Official leaderboard |
| 30 | GPT-6 Luna | 30% ±9 | tolokaforge | Sep 2026 | Official leaderboard |
| 31 | | 29% ±8.9 | tolokaforge | Oct 2026 | Official leaderboard |
| 32 | | 26% ±8.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 33 | | 22% ±8.2 | tolokaforge | Oct 2026 | Official leaderboard |
| 34 | Claude Haiku 5.5 | 21% ±8 | tolokaforge | Oct 2026 | Official leaderboard |
| 35 | Claude Sonnet 5 | 19% ±7.7 | tolokaforge | Jun 2026 | Official leaderboard |
| 36 | | 18% ±7.5 | tolokaforge | Oct 2026 | Official leaderboard |
| 37 | | 12% ±6.4 | tolokaforge | Nov 2026 | Official leaderboard |
| 38 | | 9.00% ±5.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 39 | | 8.00% ±5.3 | tolokaforge | Feb 2026 | Official leaderboard |
| 40 | Kimi K3 | 7.00% ±5 | tolokaforge | Jul 2026 | Official leaderboard |
| 41 | GLM-5.3 | 7.00% ±5 | tolokaforge | Aug 2026 | Official leaderboard |
| 42 | | 6.00% ±4.7 | tolokaforge | Oct 2026 | Official leaderboard |
| 43 | Flint v1.0 | 1.00% 0 to 3 | tolokaforge | Official leaderboard | |
| 44 | Mellum2.1 12B A2.5B Thinking | 0% | tolokaforge | Official leaderboard | |
| 45 | Granite 4.2 8B | 0% | tolokaforge | Aug 2026 | Official leaderboard |
| 46 | LFM2.5-24B-A2B | 0% | tolokaforge | Oct 2026 | Official leaderboard |
| # | Model | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 81% ±7.7 | tolokaforge | Nov 2026 | Official leaderboard |
| 2 | Claude Fable 5.2 | 77% ±8.2 | tolokaforge | Nov 2026 | Official leaderboard |
| 3 | | 76% ±8.4 | tolokaforge | Nov 2026 | Official leaderboard |
| 4 | Claude Opus 5.5 | 75% ±8.5 | tolokaforge | Sep 2026 | Official leaderboard |
| 5 | | 74% ±8.6 | tolokaforge | Oct 2026 | Official leaderboard |
| 6 | | 73% ±8.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 7 | GPT-6.2 Astra | 72% ±8.8 | tolokaforge | Nov 2026 | Official leaderboard |
| 8 | Claude Fable 5.1 | 71% ±8.9 | tolokaforge | Sep 2026 | Official leaderboard |
| 9 | GPT-6 Astra | 67% ±9.2 | tolokaforge | Sep 2026 | Official leaderboard |
| 10 | GPT-6.1 Sol | 66% ±9.3 | tolokaforge | Sep 2026 | Official leaderboard |
| 11 | | 65.7% ±9.4 | tolokaforge | Oct 2026 | Official leaderboard |
| 12 | | 65% ±9.3 | tolokaforge | Nov 2026 | Official leaderboard |
| 13 | GPT-6.2 Sol | 65% ±9.3 | tolokaforge | Nov 2026 | Official leaderboard |
| 14 | Claude Sonnet 5.5 | 64% ±9.4 | tolokaforge | Sep 2026 | Official leaderboard |
| 15 | Claude Opus 5 | 62% ±9.5 | tolokaforge | Jul 2026 | Official leaderboard |
| 16 | GPT-5.6 Sol | 61% ±9.6 | tolokaforge | Jul 2026 | Official leaderboard |
| 17 | | 59% ±9.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 18 | | 58% ±9.7 | tolokaforge | Aug 2026 | Official leaderboard |
| 19 | GPT-6 Sol | 58% ±9.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 20 | Grok 4.6 | 56% ±9.7 | tolokaforge | Aug 2026 | Official leaderboard |
| 21 | GPT-6.1 Terra | 55% ±9.8 | tolokaforge | Oct 2026 | Official leaderboard |
| 22 | | 52% ±9.8 | tolokaforge | Nov 2026 | Official leaderboard |
| 23 | MiMo-V2.6 Pro | 51% ±9.9 | tolokaforge | Sep 2026 | Official leaderboard |
| 24 | Kimi K3.1 Pro | 49% ±9.8 | tolokaforge | Oct 2026 | Official leaderboard |
| 25 | | 44% ±9.7 | tolokaforge | Sep 2026 | Official leaderboard |
| 26 | | 36% ±9.4 | tolokaforge | Aug 2026 | Official leaderboard |
| 27 | | 32% ±9.2 | tolokaforge | Nov 2026 | Official leaderboard |
| 28 | Grok 4.7 | 31% ±9.1 | tolokaforge | Sep 2026 | Official leaderboard |
| 29 | GPT-6 Luna | 30% ±9 | tolokaforge | Sep 2026 | Official leaderboard |
| 30 | | 29% ±8.9 | tolokaforge | Oct 2026 | Official leaderboard |
| 31 | | 26% ±8.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 32 | | 22% ±8.2 | tolokaforge | Oct 2026 | Official leaderboard |
| 33 | Claude Haiku 5.5 | 21% ±8 | tolokaforge | Oct 2026 | Official leaderboard |
| 34 | Claude Sonnet 5 | 19% ±7.7 | tolokaforge | Jun 2026 | Official leaderboard |
| 35 | | 18% ±7.5 | tolokaforge | Oct 2026 | Official leaderboard |
| 36 | | 12% ±6.4 | tolokaforge | Nov 2026 | Official leaderboard |
| 37 | | 9.00% ±5.6 | tolokaforge | Sep 2026 | Official leaderboard |
| 38 | | 8.00% ±5.3 | tolokaforge | Feb 2026 | Official leaderboard |
| 39 | Kimi K3 | 7.00% ±5 | tolokaforge | Jul 2026 | Official leaderboard |
| 40 | GLM-5.3 | 7.00% ±5 | tolokaforge | Aug 2026 | Official leaderboard |
| 41 | | 6.00% ±4.7 | tolokaforge | Oct 2026 | Official leaderboard |
| 42 | Granite 4.2 8B | 0% | tolokaforge | Aug 2026 | Official leaderboard |
| 43 | LFM2.5-24B-A2B | 0% | tolokaforge | Oct 2026 | Official leaderboard |
Performance vs. cost
The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.
| Model | Setting | Reliability score (pass^5) | Average cost per task |
|---|---|---|---|
| tolokaforge | 81% | $4.55 | |
| Claude Fable 5.2 | tolokaforge | 77% | $15.93 |
| tolokaforge | 76% | $2.01 | |
| Claude Opus 5.5 | tolokaforge | 75% | $8.46 |
| tolokaforge | 74% | $4.5 | |
| tolokaforge | 73% | $2.41 | |
| GPT-6.2 Astra | tolokaforge | 72% | $8.22 |
| Claude Fable 5.1 | tolokaforge | 71% | $21.23 |
| GPT-6 Astra | tolokaforge | 67% | $12.8 |
| GPT-6.1 Sol | tolokaforge | 66% | $2.49 |
| tolokaforge | 65.7% | $7.26 | |
| tolokaforge | 65% | $2.2 | |
| GPT-6.2 Sol | tolokaforge | 65% | $2.63 |
| Claude Sonnet 5.5 | tolokaforge | 64% | $4.11 |
| Claude Opus 5 | tolokaforge | 62% | $11.26 |
| GPT-5.6 Sol | tolokaforge | 61% | $2.57 |
| tolokaforge | 59% | $1.87 | |
| tolokaforge | 58% | $1.1 | |
| GPT-6 Sol | tolokaforge | 58% | $3.21 |
| Grok 4.6 | tolokaforge | 56% | $4.13 |
| GPT-6.1 Terra | tolokaforge | 55% | $1.48 |
| tolokaforge | 52% | $0.16 | |
| MiMo-V2.6 Pro | tolokaforge | 51% | $0.75 |
| Kimi K3.1 Pro | tolokaforge | 49% | $2.62 |
| tolokaforge | 44% | $2.75 | |
| Reflection Beam | tolokaforge | 40% | $0.32 |
| tolokaforge | 36% | $0.65 | |
| tolokaforge | 32% | $0.98 | |
| Grok 4.7 | tolokaforge | 31% | $2.2 |
| GPT-6 Luna | tolokaforge | 30% | $0.15 |
| tolokaforge | 29% | $1.1 | |
| tolokaforge | 26% | $0.45 | |
| tolokaforge | 22% | $0.33 | |
| Claude Haiku 5.5 | tolokaforge | 21% | $0.21 |
| Claude Sonnet 5 | tolokaforge | 19% | $3.54 |
| tolokaforge | 18% | $2.25 | |
| tolokaforge | 12% | $0.62 | |
| tolokaforge | 9% | $1.51 | |
| tolokaforge | 8% | $1.69 | |
| GLM-5.3 | tolokaforge | 7% | $1.54 |
| Kimi K3 | tolokaforge | 7% | $3.17 |
| tolokaforge | 6% | $0.15 | |
| Flint v1.0 | tolokaforge | 1% | $0.12 |
| Granite 4.2 8B | tolokaforge | 0% | $0.08 |
| LFM2.5-24B-A2B | tolokaforge | 0% | $0.1 |
| Mellum2.1 12B A2.5B Thinking | tolokaforge | 0% | $0.12 |
Performance vs. cost
The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.
| Model | Setting | Reliability score (pass^5) | Average cost per task |
|---|---|---|---|
| tolokaforge | 81% | $4.55 | |
| Claude Fable 5.2 | tolokaforge | 77% | $15.93 |
| tolokaforge | 76% | $2.01 | |
| Claude Opus 5.5 | tolokaforge | 75% | $8.46 |
| tolokaforge | 74% | $4.5 | |
| tolokaforge | 73% | $2.41 | |
| GPT-6.2 Astra | tolokaforge | 72% | $8.22 |
| Claude Fable 5.1 | tolokaforge | 71% | $21.23 |
| GPT-6 Astra | tolokaforge | 67% | $12.8 |
| GPT-6.1 Sol | tolokaforge | 66% | $2.49 |
| tolokaforge | 65.7% | $7.26 | |
| tolokaforge | 65% | $2.2 | |
| GPT-6.2 Sol | tolokaforge | 65% | $2.63 |
| Claude Sonnet 5.5 | tolokaforge | 64% | $4.11 |
| Claude Opus 5 | tolokaforge | 62% | $11.26 |
| GPT-5.6 Sol | tolokaforge | 61% | $2.57 |
| tolokaforge | 59% | $1.87 | |
| tolokaforge | 58% | $1.1 | |
| GPT-6 Sol | tolokaforge | 58% | $3.21 |
| Grok 4.6 | tolokaforge | 56% | $4.13 |
| GPT-6.1 Terra | tolokaforge | 55% | $1.48 |
| tolokaforge | 52% | $0.16 | |
| MiMo-V2.6 Pro | tolokaforge | 51% | $0.75 |
| Kimi K3.1 Pro | tolokaforge | 49% | $2.62 |
| tolokaforge | 44% | $2.75 | |
| tolokaforge | 36% | $0.65 | |
| tolokaforge | 32% | $0.98 | |
| Grok 4.7 | tolokaforge | 31% | $2.2 |
| GPT-6 Luna | tolokaforge | 30% | $0.15 |
| tolokaforge | 29% | $1.1 | |
| tolokaforge | 26% | $0.45 | |
| tolokaforge | 22% | $0.33 | |
| Claude Haiku 5.5 | tolokaforge | 21% | $0.21 |
| Claude Sonnet 5 | tolokaforge | 19% | $3.54 |
| tolokaforge | 18% | $2.25 | |
| tolokaforge | 12% | $0.62 | |
| tolokaforge | 9% | $1.51 | |
| tolokaforge | 8% | $1.69 | |
| GLM-5.3 | tolokaforge | 7% | $1.54 |
| Kimi K3 | tolokaforge | 7% | $3.17 |
| tolokaforge | 6% | $0.15 | |
| Granite 4.2 8B | tolokaforge | 0% | $0.08 |
| LFM2.5-24B-A2B | tolokaforge | 0% | $0.1 |
The best score, 81%, costs $4.54 a task
Runs on the cost frontier score higher than any run that's cheaper. Gemini 4 Hydrogen comes within 5 points of the best for $2.01 a task, about half the cost. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official Toloka Arena Enterprise Tool Use (November 2026) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| tolokaforge | 81% | $4.54 | |||
| Claude Fable 5.2 | tolokaforge | 77% | $16 | ||
| tolokaforge | 76% | $2.01 | |||
| Claude Opus 5.5 | tolokaforge | 75% | $8.46 | ||
| tolokaforge | 74% | $4.50 | |||
| tolokaforge | 73% | $2.41 | |||
| GPT-6.2 Astra | tolokaforge | 72% | $8.22 | ||
| Claude Fable 5.1 | tolokaforge | 71% | $21 | ||
| GPT-6 Astra | tolokaforge | 67% | $13 | ||
| GPT-6.1 Sol | tolokaforge | 66% | $2.49 | ||
| tolokaforge | 65.7% | $7.26 | |||
| tolokaforge | 65% | $2.20 | |||
| GPT-6.2 Sol | tolokaforge | 65% | $2.63 | ||
| Claude Sonnet 5.5 | tolokaforge | 64% | $4.11 | ||
| Claude Opus 5 | tolokaforge | 62% | $11 | ||
| GPT-5.6 Sol | tolokaforge | 61% | $2.57 | ||
| tolokaforge | 59% | $1.87 | |||
| tolokaforge | 58% | $1.10 | |||
| GPT-6 Sol | tolokaforge | 58% | $3.21 | ||
| Grok 4.6 | tolokaforge | 56% | $4.13 | ||
| GPT-6.1 Terra | tolokaforge | 55% | $1.48 | ||
| tolokaforge | 52% | $0.16 | |||
| MiMo-V2.6 Pro | tolokaforge | 51% | $0.75 | ||
| Kimi K3.1 Pro | tolokaforge | 49% | $2.62 | ||
| tolokaforge | 44% | $2.75 | |||
| tolokaforge | 36% | $0.65 | |||
| tolokaforge | 32% | $0.98 | |||
| Grok 4.7 | tolokaforge | 31% | $2.20 | ||
| GPT-6 Luna | tolokaforge | 30% | $0.15 | ||
| tolokaforge | 29% | $1.09 | |||
| tolokaforge | 26% | $0.45 | |||
| tolokaforge | 22% | $0.33 | |||
| Claude Haiku 5.5 | tolokaforge | 21% | $0.21 | ||
| Claude Sonnet 5 | tolokaforge | 19% | $3.54 | ||
| tolokaforge | 18% | $2.25 | |||
| tolokaforge | 12% | $0.62 | |||
| tolokaforge | 9.00% | $1.51 | |||
| tolokaforge | 8.00% | $1.69 | |||
| Kimi K3 | tolokaforge | 7.00% | $3.17 | ||
| GLM-5.3 | tolokaforge | 7.00% | $1.54 | ||
| tolokaforge | 6.00% | $0.15 | |||
| Granite 4.2 8B | tolokaforge | 0% | $0.08 | ||
| LFM2.5-24B-A2B | tolokaforge | 0% | $0.10 |
The best score, 81%, costs $4.54 a task
Runs on the cost frontier score higher than any run that's cheaper. Gemini 4 Hydrogen comes within 5 points of the best for $2.01 a task, about half the cost. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official Toloka Arena Enterprise Tool Use (November 2026) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| tolokaforge | 81% | $4.54 | |||
| Claude Fable 5.2 | tolokaforge | 77% | $16 | ||
| tolokaforge | 76% | $2.01 | |||
| Claude Opus 5.5 | tolokaforge | 75% | $8.46 | ||
| tolokaforge | 74% | $4.50 | |||
| tolokaforge | 73% | $2.41 | |||
| GPT-6.2 Astra | tolokaforge | 72% | $8.22 | ||
| Claude Fable 5.1 | tolokaforge | 71% | $21 | ||
| GPT-6 Astra | tolokaforge | 67% | $13 | ||
| GPT-6.1 Sol | tolokaforge | 66% | $2.49 | ||
| tolokaforge | 65.7% | $7.26 | |||
| tolokaforge | 65% | $2.20 | |||
| GPT-6.2 Sol | tolokaforge | 65% | $2.63 | ||
| Claude Sonnet 5.5 | tolokaforge | 64% | $4.11 | ||
| Claude Opus 5 | tolokaforge | 62% | $11 | ||
| GPT-5.6 Sol | tolokaforge | 61% | $2.57 | ||
| tolokaforge | 59% | $1.87 | |||
| tolokaforge | 58% | $1.10 | |||
| GPT-6 Sol | tolokaforge | 58% | $3.21 | ||
| Grok 4.6 | tolokaforge | 56% | $4.13 | ||
| GPT-6.1 Terra | tolokaforge | 55% | $1.48 | ||
| tolokaforge | 52% | $0.16 | |||
| MiMo-V2.6 Pro | tolokaforge | 51% | $0.75 | ||
| Kimi K3.1 Pro | tolokaforge | 49% | $2.62 | ||
| tolokaforge | 44% | $2.75 | |||
| tolokaforge | 36% | $0.65 | |||
| tolokaforge | 32% | $0.98 | |||
| Grok 4.7 | tolokaforge | 31% | $2.20 | ||
| GPT-6 Luna | tolokaforge | 30% | $0.15 | ||
| tolokaforge | 29% | $1.09 | |||
| tolokaforge | 26% | $0.45 | |||
| tolokaforge | 22% | $0.33 | |||
| Claude Haiku 5.5 | tolokaforge | 21% | $0.21 | ||
| Claude Sonnet 5 | tolokaforge | 19% | $3.54 | ||
| tolokaforge | 18% | $2.25 | |||
| tolokaforge | 12% | $0.62 | |||
| tolokaforge | 9.00% | $1.51 | |||
| tolokaforge | 8.00% | $1.69 | |||
| Kimi K3 | tolokaforge | 7.00% | $3.17 | ||
| GLM-5.3 | tolokaforge | 7.00% | $1.54 | ||
| tolokaforge | 6.00% | $0.15 | |||
| Granite 4.2 8B | tolokaforge | 0% | $0.08 | ||
| LFM2.5-24B-A2B | tolokaforge | 0% | $0.10 |
3 models have held number one on the official leaderboard since its release in November 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Toloka Arena Enterprise Tool Use (November 2026) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Toloka Arena Enterprise Tool Use (November 2026) 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
- 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
All 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access The tasks are private.
- Pass Confidence Checked against a primary source.
Scores 5 pass
- Pass Independent runs Scores from the official leaderboard (46 entries).
- Pass Models scored 43 models we track, plus 3 other entries on its leaderboard.
- Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
- Pass Frontier models 10 of the top 10 have a score, led by GPT-6.2 Astra (#1).
- Pass Headroom The best is 81%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on toloka.ai: 43 rows for models we track and 3 other entries at the last read.
- Pass Last read November 20th, 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.