Autonomous Agent Backbone Benchmarks · 2026

Agentic Foundation Models
Leaderboard & Arena

Independent evaluations on SWE-bench Verified, Berkeley Function-Calling Leaderboard (BFCL), multi-hop reasoning, and real-world MCP tool-use latency across frontier autonomous models.

SWE-bench Verified SOTABFCL Function CallingAnthropic MCP Native

Frontier Agent Champions

Ranked by Agent Arena ELO
🥇Rank #1
Coding SOTA #1

Claude Fable 5.1

by Anthropic

1520Agent Arena Elo

Anthropic's 2026 flagship autonomous agent model. Tops the LMSYS Arena with an unprecedented 1570 Coding Elo, native deep multi-turn planning, and full-spectrum OS terminal execution.

1M tokens · ~95 tps
Inspect
🥈Rank #2
Co-Champion Frontier

GPT-6 Astra

by OpenAI

1520Agent Arena Elo

OpenAI's next-generation reasoning foundation architecture. Tied for #1 in LMSYS Arena general text with 1520 Elo, featuring autonomous test-time compute scaling and real-time self-verification.

1M tokens · ~120 tps
Inspect
🥉Rank #3
Supreme Architecture

Claude Opus 5

by Anthropic

1511Agent Arena Elo

Anthropic's maximum intelligence tier for complex software architecture redesign, formal verification, and multi-disciplinary deep research swarms.

1M tokens · ~70 tps
Inspect
RankModel & CreatorArena EloSWE-bench VerifiedContext & VelocityPricing & ProtocolActions
#1+1
Claude Fable 5.1Coding SOTA #1 CommercialMCP Native

by Anthropic

1520Elo
78.4%

Verified Solve

1M tokens
Velocity: ~95 tps
Pay per 1M Tokens
$10.00 input / $50.00 output per 1M
Specs
#2+2
GPT-6 AstraCo-Champion Frontier Commercial

by OpenAI

1520Elo
77.2%

Verified Solve

1M tokens
Velocity: ~120 tps
Pay per 1M Tokens
$2.50 input / $12.50 output per 1M
Specs
#3+0
Claude Opus 5Supreme Architecture CommercialMCP Native

by Anthropic

1511Elo
76.8%

Verified Solve

1M tokens
Velocity: ~70 tps
Pay per 1M Tokens
$15.00 input / $75.00 output per 1M
Specs
#4+0
GPT-5.6 SolMath & Code Elite Commercial

by OpenAI

1509Elo
75.1%

Verified Solve

1M tokens
Velocity: ~115 tps
Pay per 1M Tokens
$1.80 input / $7.20 output per 1M
Specs
#5+2
Grok 4.6Supercluster SOTA Commercial

by xAI

1507Elo
73.9%

Verified Solve

500K tokens
Velocity: ~140 tps
Pay per 1M Tokens
$2.00 input / $8.00 output per 1M
Specs
#6+4
Kimi-K3Open Weights Champion Open WeightsMCP Native

by Moonshot AI

1506Elo
74.2%

Verified Solve

1M tokens
Velocity: ~85 tps
Open Weights (Self-hosted)
Open weights on HuggingFace / API: $0.60 per 1M
Specs
#7+2
Qwen 3.8 MaxOpen Weights SOTA Open WeightsMCP Native

by Alibaba Cloud Qwen

1506Elo
73.5%

Verified Solve

1M tokens
Velocity: ~105 tps
Open Weights (Self-hosted)
Apache 2.0 on HuggingFace / Bailian API $0.40/1M
Specs
#8+1
Gemini 3.8 FlashSpeed & Scale SOTA Commercial

by Google DeepMind

1505Elo
72.8%

Verified Solve

1M tokens
Velocity: ~210 tps
Pay per 1M Tokens
$0.30 input / $1.20 output per 1M (Free tier available)
Specs
#9+1
GLM-5.3MIT Open Champion Open WeightsMCP Native

by Z.ai (Zhipu)

1505Elo
72.1%

Verified Solve

1M tokens
Velocity: ~90 tps
Open Weights (Self-hosted)
MIT Open Source / Zhipu BigModel API
Specs
#10+3
DeepSeek-V4-ProExtreme Value SOTA Open Weights

by DeepSeek AI

1502Elo
71.8%

Verified Solve

1M tokens
Velocity: ~80 tps
Open Weights (Self-hosted)
Open weights on HF / $0.25 input / $0.50 output per 1M
Specs
#11+0
Claude Haiku 4.5Sub-Second Router CommercialMCP Native

by Anthropic

1485Elo
68.5%

Verified Solve

200K tokens
Velocity: ~190 tps
Pay per 1M Tokens
$0.80 input / $4.00 output per 1M
Specs