| Provider | Anthropic |
| Released | 2024-03-04 |
| Status | Deprecated |
| Context window | 200K tokens |
| Pricing | Paid |
| Input price | $15.00/M |
| Output price | $75.00/M |
| Capabilities | textvisioncodefunction-calling |
| API docs | Docs ↗ |
| Replaced by | claude-opus-4-6 |
As of 2024-12-01. Source
| Claude 3 Opus | Mistral Large 2 | Gemini 2.5 Pro | Claude Opus 4.6 | |
|---|---|---|---|---|
| Context | 200K | 131K | 1.0M | 200K |
| MMLU | 86.8 | 84.0 | 91.0 | 92.0 |
| HumanEval | 84.9 | 92.0 | — | 94.5 |
| MATH | 60.1 | 69.7 | 97.0 | 97.1 |
| GPQA | 50.4 | — | 84.0 | 79.9 |
| Pricing | Paid | Paid | Paid | Paid |
| Input $/M | $15.00 | $2.00 | $1.25 | $15.00 |
Claude 3 Opus was Anthropic's most intelligent model at launch in March 2024, praised for nuanced instruction-following and complex reasoning. It has been superseded by Claude 3.5 Sonnet and Claude Opus 4.6.
For new projects requiring maximum Anthropic intelligence, use Claude Opus 4.6. For a balance of performance and cost, use Claude 3.5 Sonnet.