The agents leaderboard · for developers

Best for agents

4 editor picks · 7 eligible models · Tool-use loops that don't crash on turn 12.

See raw /best

EDITOR'S CHOICEResearched 1d ago

Claude Sonnet 4.6

Anthropic · 1M context

Excellent

The most reliable tool-use loop in production — recovers from errors on its own.

Best generally-available τ-bench (87.5); stays on-task across long tool loops and self-corrects without prompting.

Open model

The numbers

$/1M out

$15.00

$3.00 input

Context

max window

Pros

+Top GA τ-bench score
+Reliable multi-step recovery
+1M context

Cons

−$15 / 1M out
−Not the cheapest per step

Also worth picking

The runners-up

ranked by editorial pick order

Editorial tiersExcellentStrongSolid

#ModelTier$/1M outEditor's note

GLM-5

Zhipu AI · 200K

$2.08

τ-bench 82.1, open weights at $2.08 out — the best price-per-step agent we'd run always-on.

GPT-5.4

OpenAI · 1M

$15.00

Picked over GPT-5.5 here: it carries the measured τ-bench score (78.3) and runs at half the output price ($15 vs $30) — decisive for always-on agents.

Gemini 3 Pro

Google DeepMind · 1M

$5.00

Best for chatty agents with many read-only tools and large contexts.

Eligibility

7 models are eligible for this board

Eligibility means tagged with useCases: [agents]. Pins must come from this pool.

All picks