LLM Reference

The long context leaderboard · for developers

Best for long context

5 editor picks · 8 eligible models · Past 200K tokens without falling apart.

Editorial pick plus benchmark and API pricing context.

See raw /best
EDITOR'S CHOICEResearched 186d ago

Gemini 3 Pro

Google DeepMind · 1m context
Excellent

The models still credible when the window is genuinely full.

The long-context standard — 1M tokens with cheap input that keeps the bill survivable.

Open model
The numbers
$/1M out
$5.00
$1.25 input
Context
1m
max window
Pros
  • +1M context, strong recall
  • +Cheap input tokens
  • +Mature tooling
Cons
  • −Not the strongest pure reasoner

Also worth picking

The runners-up

ranked by editorial pick orderEditorial tiersExcellentStrongSolid
#2
Claude Sonnet 5
Anthropic · 1m
$10.00 / 1M out
Agentic long-context pick: 1M window with strong BrowseComp and OSWorld scores — use when the job is tool loops across a full corpus, not cheapest input tokens.
#3
Claude Opus 4.7
Anthropic · 1m
$25.00 / 1M out
Catches subtle cross-references across a full 1M window better than anyone; pricey but precise.
#4
GPT-5.4
OpenAI · 1.05m
$15.00 / 1M out
Same 1.05M window as GPT-5.5 at half the output price ($15 vs $30) — the value choice for long analytical passes.
#5
DeepSeek V4 Flash
DeepSeek · 1m
$0.18 / 1M out
1M context at $0.22 out, open weights — the value pick for huge inputs.

Eligibility

8 models are eligible for this board

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

All picks