Long-context leaderboard
LLMs with the largest context windows
Ranked by maximum input tokens the model accepts in a single call. Larger windows unlock whole-codebase analysis, long-document summarization and multi-hour transcripts — though effective recall often lags the advertised window.
Category leader
#01OpenAI
OpenAI: GPT-5.6 Sol.
GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series. It is suited for complex reasoning, coding, and agentic workflows, and is particularly strong at command-line and multi-step coding tasks...
Context
1.1M
tokens
Ranking
N=20| # | Model | Provider | Context | Ctx | Released |
|---|---|---|---|---|---|
| 01 | OpenAI: GPT-5.6 Sol | OpenAI | 1.1M | 1.1M | 2026-07-09 |
| 02 | OpenAI: GPT-5.6 Terra | OpenAI | 1.1M | 1.1M | 2026-07-09 |
| 03 | OpenAI: GPT-5.6 Terra Pro | OpenAI | 1.1M | 1.1M | 2026-07-09 |
| 04 | MoonshotAI: Kimi K3 | Moonshot | 1.0M | 1.0M | 2026-07-16 |
| 05 | DeepSeek: DeepSeek V4 Pro | DeepSeek | 1.0M | 1.0M | 2026-04-24 |
| 06 | Z.ai: GLM 5.2 | Z.ai | 1.0M | 1.0M | 2026-06-16 |
| 07 | Meta: Muse Spark 1.1 | Meta | 1.0M | 1.0M | 2026-07-16 |
| 08 | MiniMax: MiniMax M3 | MiniMax | 1.0M | 1.0M | 2026-05-31 |
| 09 | Anthropic: Claude Fable 5 | Anthropic | 1.0M | 1.0M | 2026-06-09 |
| 10 | Qwen: Qwen3.7 Max | Alibaba | 1.0M | 1.0M | 2026-05-21 |
| 11 | Anthropic: Claude Opus 4.8 | Anthropic | 1.0M | 1.0M | 2026-05-27 |
| 12 | Anthropic: Claude Opus 4.8 (Fast) | Anthropic | 1.0M | 1.0M | 2026-05-27 |
| 13 | NVIDIA: Nemotron 3 Ultra | NVIDIA | 1.0M | 1.0M | 2026-06-04 |
| 14 | Sakana: Fugu Ultra | Sakana | 1.0M | 1.0M | 2026-06-24 |
| 15 | xAI: Grok 4.3 | xAI | 1.0M | 1.0M | 2026-04-30 |
| 16 | xAI: Grok 4.5 | xAI | 500K | 500K | 2026-07-08 |
| 17 | MoonshotAI: Kimi K2.7 Code | Moonshot | 262K | 262K | 2026-06-12 |
| 18 | MoonshotAI: Kimi K2.6 | Moonshot | 262K | 262K | 2026-04-20 |
| 19 | Mistral: Mistral Medium 3.5 | Mistral | 262K | 262K | 2026-04-30 |
| 20 | Z.ai: GLM 5.1 | Z.ai | 203K | 203K | 2026-04-07 |
Frequently asked
- Which LLM has the largest context window?
- The current leader is at the top of this table. Context windows range from 128K to 2M+ tokens across frontier models; larger windows unlock whole-codebase and full-document workflows.
- Is bigger context always better?
- Not necessarily. Recall quality often degrades past ~200K tokens even when the window advertises 1M+. Use benchmarks like Needle-in-a-Haystack to verify effective context, not just declared window size.