Which AI model or agent should you use?

Live standings for 72 frontier models and 12 agent frameworks β€” capability measured on the kind of work you're actually doing, priced against what it costs per million tokens. Every number links to its public source, and anything we can't source is left blank instead of estimated.

Auto-refreshed daily Β· last update 2026-11-22 05:43 UTC

Straight answers

The picks below come from measured human-preference scores on each kind of work, divided against live API pricing. Change the task or set a budget β†’

Best model for coding

Writing and editing code
Gemini 4 Carbon
Google Β· $4.00/1M blended
Nothing cheaper stays within 90% of it on this board.

Best model for agents

Autonomous multi-step tool use
Claude Fable 5.2
Anthropic Β· $16.00/1M blended
Nothing cheaper stays within 90% of it on this board.

Best all-round model

General assistant work
Gemini 4 Carbon
Google Β· $4.00/1M blended
Nothing cheaper stays within 90% of it on this board.

New releases last 3 weeks

ModelProviderReleasedContextInput /1MOutput /1M
Pixtral Large 2502Mistral2026-11-20 new128K$2.00$6.00
Gemini 4 NeonGoogle2026-11-18 new1.048576M$3.00$15.00
GPT-6.2 Astra ProOpenAI2026-11-17 new1.05M$8.00$40.00
GPT-6.2 AstraOpenAI2026-11-17 new1.05M$8.00$40.00
GPT-6.2 Sol ProOpenAI2026-11-17 new1.05M$2.00$10.00
GPT-6.2 SolOpenAI2026-11-17 new1.05M$2.00$10.00
Qwen4 MaxQwen (Alibaba)2026-11-16 new1M$1.42$4.27
Gemini 4 HydrogenGoogle2026-11-12 new1.048576M$1.50$7.50
Gemini 4 CarbonGoogle2026-11-10 new1.048576M$2.00$10.00
Claude Fable 5.2Anthropic2026-11-09 new1M$8.00$40.00
Gemini Omni 1.2 FlashGoogle2026-11-05 new131K$1.50$9.00
Gemini Omni 1.1 FlashGoogle2026-11-04 new131K$1.50$9.00
GLM 5.3 FlashX LegalZ.ai (Zhipu)2026-11-02 new1.048576M$0.37$1.25

Source: OpenRouter model catalog Β· retrieved 2026-11-22

All tracked models 72

General and Coding are Arena Elo for that category β€” human preference votes, higher is better. AA is the Artificial Analysis Intelligence Index, a benchmark-based composite on a different scale; it's shown beside the Elo columns as an independent cross-check, never blended with them. Blank means no published score, never an estimate. Price is blended (inputΓ—3 + outputΓ—1) Γ· 4 per million tokens; 18 models tagged varies are priced differently by UTC day or hour, and the full schedule is in the API. Click any header to sort.

ModelProviderGeneralCodingAA$/1MContextToolsReleased
Pixtral Large 2502
mistralai/pixtral-large-2502
Mistral β€” β€” β€” $3.00 128K β€” 2026-11-20
Gemini 4 Neon
google/gemini-4-neon
Google 1524 1555 61.4 $6.00 1.048576M tools 2026-11-18
GPT-6.2 Astra Pro
openai/gpt-6.2-astra-pro
OpenAI β€” β€” β€” $16.00 varies 1.05M tools 2026-11-17
GPT-6.2 Astra
openai/gpt-6.2-astra
OpenAI 1414 1450 62.1 $16.00 varies 1.05M tools 2026-11-17
GPT-6.2 Sol Pro
openai/gpt-6.2-sol-pro
OpenAI β€” β€” β€” $4.00 varies 1.05M tools 2026-11-17
GPT-6.2 Sol
openai/gpt-6.2-sol
OpenAI 1414 1440 57.3 $4.00 varies 1.05M tools 2026-11-17
Qwen4 Max
qwen/qwen4-max
Qwen (Alibaba) β€” β€” 42.9 $2.13 1M tools 2026-11-16
Gemini 4 Hydrogen
google/gemini-4-hydrogen
Google 1502 1512 53.9 $3.00 1.048576M tools 2026-11-12
Gemini 4 Carbon
google/gemini-4-carbon
Google 1539 1564 60 $4.00 1.048576M tools 2026-11-10
Claude Fable 5.2
anthropic/claude-fable-5.2
Anthropic 1478 1487 64.6 $16.00 1M tools 2026-11-09
Gemini Omni 1.2 Flash
google/gemini-omni-1.2-flash
Google β€” β€” β€” $3.38 131K β€” 2026-11-05
Gemini Omni 1.1 Flash
google/gemini-omni-1.1-flash
Google β€” β€” β€” $3.38 131K β€” 2026-11-04
GLM 5.3 FlashX Legal
z-ai/glm-5.3-flashx-legal
Z.ai (Zhipu) β€” β€” β€” $0.59 1.048576M tools 2026-11-02
GPT-6.1 Terra Pro
openai/gpt-6.1-terra-pro
OpenAI β€” β€” β€” $2.00 varies 1.05M tools 2026-10-26
GPT-6.1 Terra
openai/gpt-6.1-terra
OpenAI 1426 1482 51.3 $2.00 varies 1.05M tools 2026-10-26
Kimi K3.1 Pro
moonshotai/kimi-k3.1-pro
Moonshot AI 1458 1484 46.5 $2.39 1.048576M tools 2026-10-23
MiniMax M3.1 Flash
minimax/minimax-m3.1-flash
MiniMax β€” β€” 29.1 $0.26 varies 1.048576M tools 2026-10-19
DeepSeek V4.1 Flash 1017
deepseek/deepseek-v4.1-flash-1017
DeepSeek β€” β€” 32.6 $0.26 varies 1.048576M tools 2026-10-17
Gemini 4 Argon
google/gemini-4-argon
Google 1528 1539 50.4 $4.00 1.048576M tools 2026-10-13
Claude Haiku 5.5
anthropic/claude-haiku-5.5
Anthropic 1398 1452 36.5 $0.20 varies 1M tools 2026-10-07
Mistral Large 4
mistralai/mistral-large-4-0
Mistral β€” β€” 26.1 $1.05 1.048576M tools 2026-10-06
Nano Banana 2.1
google/gemini-nano-banana-2.1
Google β€” β€” β€” $3.00 66K tools 2026-10-06
GPT-6.1 Sol Pro
openai/gpt-6.1-sol-pro
OpenAI β€” β€” β€” $4.00 varies 1.05M tools 2026-09-29
GPT-6.1 Sol
openai/gpt-6.1-sol
OpenAI 1446 1490 54.6 $4.00 varies 1.05M tools 2026-09-29
Claude Sonnet 5.5
anthropic/claude-sonnet-5.5
Anthropic 1470 1500 54.5 $4.00 1M tools 2026-09-28
Qwen3.8 Max Prime
qwen/qwen3.8-max-prime
Qwen (Alibaba) β€” β€” β€” $6.00 1M tools 2026-09-23
GLM 5.3 Prime
z-ai/glm-5.3-prime
Z.ai (Zhipu) β€” β€” β€” $2.58 1M tools 2026-09-23
Claude Opus 5.5
anthropic/claude-opus-5.5
Anthropic 1513 1543 59.1 $8.00 1M tools 2026-09-22
Command A+
cohere/command-a-plus
Cohere β€” β€” 6.2 $0.60 192K tools 2026-09-22
Qwen3.8 Omni Flash
qwen/qwen3.8-omni-flash
Qwen (Alibaba) β€” β€” β€” $0.23 1M tools 2026-09-21
Grok 4.7
x-ai/grok-4.7
xAI 1400 1423 35.5 $3.00 varies 500K tools 2026-09-21
GLM 5.3 FlashX
z-ai/glm-5.3-flashx
Z.ai (Zhipu) β€” β€” β€” $0.59 1.048576M tools 2026-09-18
DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flash
DeepSeek 1463 1503 β€” $0.09 1.048576M tools 2026-09-10
Qwen3.8 Max (0902)
qwen/qwen3.8-max-0902
Qwen (Alibaba) 1485 1503 33.7 $3.00 1M tools 2026-09-03
Gemini 3.8 Flash
google/gemini-3.8-flash
Google 1500 1509 30.8 $1.66 1.048576M tools 2026-09-02
Claude Fable 5.1
anthropic/claude-fable-5.1
Anthropic 1512 1527 50 $20.00 1M tools 2026-09-01
Qwen3.8 Flash
qwen/qwen3.8-flash
Qwen (Alibaba) β€” β€” 28.6 $0.23 1M tools 2026-08-26
GLM 5.3 Flash
z-ai/glm-5.3-flash
Z.ai (Zhipu) 1471 1509 27.8 $0.05 1.048576M tools 2026-08-26
DeepSeek V4 Flash Vision Exp
deepseek/deepseek-v4-flash-vision-exp
DeepSeek β€” β€” β€” $0.33 1.048576M tools 2026-08-21
GLM 5.3
z-ai/glm-5.3
Z.ai (Zhipu) 1475 1501 31.7 $0.40 1.048576M tools 2026-08-18
Qwen3.8 27B
qwen/qwen3.8-27b
Qwen (Alibaba) 1437 1481 20.1 $0.41 1.024M tools 2026-08-14
Qwen3.8 2.4T A95B
qwen/qwen3.8-2.4t-a95b
Qwen (Alibaba) β€” β€” β€” $3.00 1.048576M tools 2026-08-12
DeepSeek V4 Pro 0813
deepseek/deepseek-v4-pro-0813
DeepSeek 1451 1469 23.1 $1.98 1.048576M tools 2026-08-12
Grok 4.6
x-ai/grok-4.6
xAI 1430 1464 31 $3.00 varies 500K tools 2026-08-12
DeepSeek V4 Flash 0731
deepseek/deepseek-v4-flash-0731
DeepSeek 1432 1457 β€” $0.10 1.048576M tools 2026-07-31
Qwen3.7 Flash
qwen/qwen3.7-flash
Qwen (Alibaba) β€” β€” β€” $0.06 varies 1M tools 2026-07-27
Claude Opus 5
anthropic/claude-opus-5
Anthropic 1506 1532 44.2 $10.00 1M tools 2026-07-24
Kimi K3
moonshotai/kimi-k3
Moonshot AI 1475 1505 17.4 $4.14 1.048576M tools 2026-07-16
Grok 4.5
x-ai/grok-4.5
xAI 1448 1474 β€” $3.00 varies 500K tools 2026-07-08
Claude Sonnet 5
anthropic/claude-sonnet-5
Anthropic 1442 1483 β€” $4.00 1M tools 2026-06-30
GLM 5.2
z-ai/glm-5.2
Z.ai (Zhipu) 1469 1488 β€” $1.07 1.048576M tools 2026-06-16
Kimi K2.7 Code
moonshotai/kimi-k2.7-code
Moonshot AI β€” β€” β€” $1.20 262K tools 2026-06-12
Claude Fable 5
anthropic/claude-fable-5
Anthropic 1492 1519 β€” $20.00 1M tools 2026-06-09
MiniMax M3
minimax/minimax-m3
MiniMax 1427 1467 18.9 $0.53 1.048576M tools 2026-05-31
Mistral Medium 3.5
mistralai/mistral-medium-3-5
Mistral 1421 1461 4.6 $3.00 262K tools 2026-04-30
Grok 4.3
x-ai/grok-4.3
xAI 1397 1414 β€” $1.56 varies 1M tools 2026-04-30
DeepSeek V4 Pro 0423
deepseek/deepseek-v4-pro
DeepSeek 1451 1469 20.8 $1.05 1.048576M tools 2026-04-24
DeepSeek V4 Flash 0423
deepseek/deepseek-v4-flash
DeepSeek 1432 1457 β€” $0.11 1.048576M tools 2026-04-24
Kimi K2.6
moonshotai/kimi-k2.6
Moonshot AI 1455 1489 β€” $1.34 262K tools 2026-04-20
GLM 5.1
z-ai/glm-5.1
Z.ai (Zhipu) 1463 1485 β€” $1.83 205K tools 2026-04-07
GLM 5V Turbo
z-ai/glm-5v-turbo
Z.ai (Zhipu) β€” β€” β€” $1.90 203K tools 2026-04-01
Grok 4.20 Multi-Agent
x-ai/grok-4.20-multi-agent
xAI 1451 1458 β€” $1.56 varies 2M β€” 2026-03-31
Grok 4.20
x-ai/grok-4.20
xAI 1451 1459 β€” $1.56 varies 2M tools 2026-03-31
MiniMax M2.7
minimax/minimax-m2.7
MiniMax 1405 1454 β€” $0.90 197K tools 2026-03-18
Mistral Small 4
mistralai/mistral-small-2603
Mistral β€” β€” 4.3 $0.26 262K tools 2026-03-16
MiniMax M2.5
minimax/minimax-m2.5
MiniMax 1359 1381 β€” $0.44 198K tools 2026-02-12
Kimi K2.5
moonshotai/kimi-k2.5
Moonshot AI 1445 1484 β€” $1.00 262K tools 2026-01-27
Devstral 2 2512
mistralai/devstral-2512
Mistral β€” β€” β€” $0.80 262K tools 2025-12-09
Ministral 3 14B 2512
mistralai/ministral-14b-2512
Mistral β€” β€” β€” $0.20 262K tools 2025-12-02
Ministral 3 8B 2512
mistralai/ministral-8b-2512
Mistral 1191 1202 β€” $0.15 262K tools 2025-12-02
Ministral 3 3B 2512
mistralai/ministral-3b-2512
Mistral β€” β€” β€” $0.10 131K tools 2025-12-02
DeepSeek V3.2
deepseek/deepseek-v3.2
DeepSeek 1425 1454 β€” $0.29 164K tools 2025-12-01

Pricing: OpenRouter model catalog, retrieved 2026-11-22. Elo capability: Arena leaderboard dataset (CC-BY-4.0), retrieved 2026-11-22. AA index: Artificial Analysis, republished in the OpenRouter catalog.

Top agent harnesses on SWE-bench Verified

Percentage of 500 real GitHub issues resolved. Entries are agent scaffold + model combinations, not bare models β€” the same model can move many points depending on the harness around it.

Read this board with care. We re-fetch it every day, but SWE-bench itself has not accepted a new Verified entry since 2026-02-26 β€” 269 days ago β€” so no 2026 frontier model appears here. Since late 2025 its submission policy has been limited to academic teams and established labs, which keeps most commercial agents off it entirely. Only 60 of 180 entries were re-run by the maintainers; the rest are submitter-reported.
#Agent + modelResolvedDateStatus
1Sonar Foundation Agent + Claude 4.5 Opus79.2%2025-12-05 self-reported
2live-SWE-agent + Claude 4.5 Opus medium (20251101)79.2%2025-12-15 self-reported
3TRAE + Doubao-Seed-Code78.8%2025-09-28 self-reported
4live-SWE-agent + Gemini 3 Pro Preview (2025-11-18)77.4%2025-11-20 self-reported
5EPAM AI/Run Developer Agent v20250719 + Claude 4 Sonnet76.8%2025-08-04 self-reported
6Atlassian Rovo Dev (2025-09-02)76.8%2025-09-02 self-reported
7Claude 4.5 Opus (high)76.8%2026-02-17 self-reported
8ACoder76.4%2025-08-19 self-reported
9Gemini 3 Flash (high)75.8%2026-02-17 self-reported
10MiniMax M2.5 (high)75.8%2026-02-17 self-reported

Source: SWE-bench leaderboard Β· fetched 2026-11-22, board last updated 2026-02-26 Β· full top 20 + framework standings β†’

πŸ¦€ Spotlight: gcrab our project

A small, safe, replayable self-hosted agent β€” single static Go binary, zero deps, hard 8K context budget, policy-gated exec, append-only event log Β· 0 stars on GitHub.

gcrab.com β†’See it in the standings

Are you an agent?

This whole site is machine-readable with no key and open CORS. Ask it a question directly: GET /api/v1/route/coding.json returns the ranked capability-vs-cost table for that task. Also /llms.txt, /leaderboard.md, and the full API index. See API & Bots.