Llama 3.1 405B Instruct Fp8
Meta
Key facts · verified Oct 29, 2026
Llama 3.1 405B Instruct Fp8 is a Meta model released on 23 Jul 2024. Its API price is $4 per million input tokens and $4 per million output tokens. Its context window is 128K tokens. It ranks #247 on LMArena's text leaderboard (Elo 1334, read 1 Nov 2026). Source: ai.meta.com.
Llama 3.1 405B Instruct FP8 is the FP8 build of Meta's Llama 3.1 405B instruction-tuned model, released on July 23, 2024 with open weights under the Llama 3.1 Community licence, so it can be self-hosted. Meta's model card lists a 128K-token context window. arena.ai lists pricing of $4 per million input tokens and $4 per million output tokens, not confirmed as first-party. On LMArena's text leaderboard: 1334 Elo (rank 247).
Pricing & specs checkedSource: vendor documentation · checked Oct 29, 2026LMArena: 1334 Elo (rank 247) · source · retrieved Nov 1, 2026
Full Specifications
$4 / $4 per million tokens | |
|---|---|
| Identity | |
| Developer | Meta |
| Released | Jul 2024 |
| Licence | Llama 3.1 Community |
| Self-hostable | Yes |
| Cost | |
| Blended $/1M tokensinput × 0.75 + output × 0.25 | $4.000 / 1M tokens |
| Input price | $4.000 / 1M tokens |
| Output price | $4.000 / 1M tokens |
| Capacity | |
| Context window | 128K tokens |
| Measured quality | |
| LMArena Elo | 1334 (checked Nov 2026) |
| LMArena rank | Rank 247 |
| Elo per dollarLMArena Elo ÷ blended $/1M tokens | 334 |