LLM Reference

The open weights leaderboard · for developers

Best for open weights

4 editor picks · 6 eligible models · Best models you can self-host today.

See raw /best
EDITOR'S CHOICEResearched today

DeepSeek V4 Pro

DeepSeek · 1M context
Excellent

DeepSeek closes the gap; the rest win on hosting choice and price.

Best open-weights model we've tested: #1 LiveCodeBench (93.5), 80.6 SWE-bench, 1M context, $0.87 out.

Open model
The numbers
$/1M out
$0.87
$0.43 input
Context
1M
max window
Pros
  • +Frontier-class coding, open weights
  • +1M context
  • +$0.87 / 1M out
Cons
  • −Heavy to self-host at full size

Also worth picking

The runners-up

ranked by editorial pick order
Editorial tiersExcellentStrongSolid
#ModelTier$/1M outEditor's note
#2
Kimi K2.6
Moonshot AI · 256K
$3.40
LiveCodeBench 89.6 and SWE-bench Pro 58.6 under a permissive license — the strongest agentic open model after DeepSeek.
#3
GLM-5.1
Zhipu AI · 200K
$3.50
SWE-bench Pro 58.4 with excellent agentic behavior; an easy self-hosted default.
#4
Qwen3.5-397B-A17B
Alibaba · 256K
$2.34
Strong reasoning + multilingual baseline; widely hosted MoE that's cheap to serve.

Eligibility

6 models are eligible for this board

Eligibility means tagged with useCases: [open-weights]. Pins must come from this pool.

All picks