AI Agents
Updated
Models that take multi-step actions with tools: running commands, using apps and finishing tasks on their own.
Agent leaderboard
57 models · ToolsGuide Score out of 1,000, where 1,000 is a flawless model
20 of 57 models in the chart · ·
All models
| Compare | # | Model | Creator | Licence | Input / 1M | Score | From votes | From tests · count |
|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) |   Anthropic | Proprietary | $4.00 | 614 | 645 | 580 · 2 | |
| 2 | Gemini 4 Carbon (High) |  Google | Proprietary | – | 599 | 599 | – | |
| 3 | Claude Sonnet 5.5 (Max) | Anthropic | Proprietary | $2.00 | 588 | 579 | 598 · 2 | |
| 4 | Gemini 4 Argon (High) | Google | Proprietary | $2.00 | 569 | 551 | 576 · 3 | |
| 5 | GPT-6.1 Sol (Max) |   OpenAI | Proprietary | $2.00 | 567 | 567 | – | |
| 6 | GPT-6 Astra (Max) | OpenAI | Proprietary | $10.00 | 563 | 571 | 560 · 4 | |
| 7 | Claude Fable 5.1 (Max) | Anthropic | Proprietary | $10.00 | 560 | 575 | 543 · 2 | |
| 8 | Muse Wonder Full (Max) |  Meta | Proprietary | – | 546 | 546 | – | |
| 9 | Kimi K3.1 Pro (Max) |   Moonshot AI | Proprietary | $2.00 | 540 | 511 | 553 · 3 | |
| 10 | GPT-6.1 Terra (Max) | OpenAI | Proprietary | $1.00 | 535 | 550 | 518 · 2 | |
| 11 | GPT-6 Sol (Max) | OpenAI | Proprietary | $2.00 | 527 | 527 | – | |
| 12 | Claude Opus 5 (Max) | Anthropic | Proprietary | $5.00 | 525 | 524 | 525 · 3 | |
| 13 | Mistral Large 4 |  Mistral | Proprietary | $0.70 | 500 | 422 | 538 · 3 | |
| 14 | GLM-5.3 (Max) |   Z.ai | MIT | $1.20 | 493 | 471 | 502 · 5 | |
| 15 | Muse Wonder (Max) | Meta | Proprietary | $4.00 | 488 | 488 | – | |
| 16 | Claude Fable 5 (High) | Anthropic | Proprietary | $10.00 | 485 | 528 | 468 · 3 | |
| 17 | DeepSeek-V4.1-Flash-20261017 (Max) |  DeepSeek | MIT | – | 485 | 485 | – | |
| 18 | GPT-5.6 Sol (xHigh) | OpenAI | Proprietary | $5.00 | 484 | 501 | 477 · 6 | |
| 19 | Grok 4.7 (xHigh) |   xAI | Proprietary | $2.00 | 483 | 483 | – | |
| 20 | GPT-6 Luna (Max) | OpenAI | Proprietary | $0.10 | 481 | 481 | – | |
| 21 | DeepSeek-V4.1-Flash-20260910 (Max) | DeepSeek | MIT | – | 479 | 479 | – | |
| 22 | MiMo-V2.6-Pro |  Xiaomi | MIT | $0.43 | 477 | 486 | 473 · 3 | |
| 23 | Claude Haiku 5.5 (High) | Anthropic | Proprietary | $0.10 | 475 | 475 | – | |
| 24 | Muse Spark 1.3 (Max) | Meta | Proprietary | $0.10 | 471 | 484 | 466 · 3 | |
| 25 | Gemini 3.8 Flash (High) | Google | Proprietary | $0.75 | 470 | 485 | 455 · 2 | |
| 26 | Grok 4.6 (xHigh) | xAI | Proprietary | $2.00 | 468 | 468 | – | |
| 27 | Kimi K3 (Max) | Moonshot AI | Kimi K3 license | $2.85 | 467 | 473 | 465 · 6 | |
| 28 | GLM-5.3-Flash | Z.ai | MIT | $0.15 | 464 | 448 | 471 · 4 | |
| 29 | DeepSeek-V4-Pro (High, 20260813) | DeepSeek | MIT | – | 462 | 462 | – | |
| 30 | MiMo-V2.6-Flash | Xiaomi | MIT | $0.14 | 462 | 472 | 458 · 3 | |
| 31 | Qwen3.8-Max |  Alibaba | Proprietary | $1.65 | 459 | 468 | 455 · 5 | |
| 32 | MiniMax-M3.1-Flash |  MiniMax | MiniMax Community License | $0.15 | 455 | 436 | 478 · 2 | |
| 33 | Hy4 preview |  Tencent | Apache 2.0 | – | 454 | 474 | 446 · 6 | |
| 34 | GPT-6 Luna (Medium) | OpenAI | Proprietary | $0.10 | 454 | 454 | – | |
| 35 | GPT-5.6 Terra (xHigh) | OpenAI | Proprietary | $2.00 | 453 | 455 | 452 · 6 | |
| 36 | Gemini 3.7 Flash (High) | Google | Proprietary | $0.75 | 452 | 455 | 451 · 3 | |
| 37 | Step 5 Preview | StepFun | Proprietary | – | 445 | 445 | – | |
| 38 | Claude Sonnet 5 (High) | Anthropic | Proprietary | $2.00 | 437 | 487 | 418 · 5 | |
| 39 | GLM-5.2 (Max) | Z.ai | MIT | $0.75 | 429 | 474 | 411 · 3 | |
| 40 | Muse Spark 1.1 | Meta | Proprietary | $1.25 | 427 | 411 | 433 · 3 | |
| 41 | Qwen3.8-Flash-Next | Alibaba | qwen-community-1.0 | – | 424 | 436 | 411 · 2 | |
| 42 | Beam 1.0 501B A23B | Reflection | Apache 2.0 | $0.40 | 423 | 430 | 420 · 6 | |
| 43 | GPT-6 Luna (Low) | OpenAI | Proprietary | $0.10 | 421 | 421 | – | |
| 44 | Qwen3.8-27B | Alibaba | Apache 2.0 | $0.40 | 416 | 433 | 397 · 2 | |
| 45 | Muse Spark 1.2 (xHigh) | Meta | Proprietary | $0.10 | 412 | 412 | – | |
| 46 | Hy3 | Tencent | Apache 2.0 | $0.14 | 409 | 414 | 407 · 6 | |
| 47 | Gemini 3.1 Pro Preview | Google | Proprietary | $2.00 | 405 | 433 | 394 · 4 | |
| 48 | MiMo-V2.5-Pro | Xiaomi | MIT | $0.43 | 403 | 403 | – | |
| 49 | Qwen3.7-Max | Alibaba | Proprietary | $1.25 | 394 | 419 | 384 · 3 | |
| 50 | MiniMax-M3 | MiniMax | MiniMax Community License | $0.28 | 392 | 395 | 391 · 3 | |
| 51 | Qwen3.7-Plus | Alibaba | Proprietary | – | 390 | 413 | 380 · 3 | |
| 52 | Inkling | Thinky | Apache 2.0 | $0.95 | 377 | 360 | 384 · 3 | |
| 53 | Inkling-Small | Thinky | Apache 2.0 | $0.30 | 374 | 353 | 383 · 5 | |
| 54 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) |  NVIDIA | OpenMDW-1.1 | – | 366 | 366 | – | |
| 55 | Nemotron 3 Ultra | NVIDIA | OpenMDW-1.1 | $0.50 | 355 | 361 | 349 · 2 | |
| 56 | Mistral Medium 3.5 | Mistral | Modified MIT | $1.50 | 351 | 380 | 320 · 2 | |
| 57 | Solar Pro 4 | Upstage | Proprietary | $0.30 | 349 | 331 | 357 · 3 |
Related benchmarks
Fixed tests with right and wrong answers, scored out of 100%.
Terminal-Bench 2.0
Getting real work done in a command-line terminal. · Top 10 of 54 · Higher is better
BrowseComp
Finding hard-to-locate facts by browsing the web. · Top 10 of 73 · Higher is better