Versions side by side — rows that differ across these models are highlighted.
| 3 models | Cl Anthropic Claude Fable 5 View → | De DeepSeek DeepSeek V4 Pro View → | Cl Anthropic Claude Opus 4.8 View → |
|---|---|---|---|
| Maker | Anthropic | DeepSeek | Anthropic |
| Tier | Flagship | Flagship | Flagship |
| Status | Active | Default | Active |
| Context window | 1M ctx | 1M ctx | 1M ctx |
| Knowledge cutoff | Jan 1, 2026 | Jan 1, 2026 | — |
| Released | Jun 9, 2026 | Apr 1, 2026 | May 28, 2026 |
| API string | claude-fable-5 |
| deepseek-v4-pro |
| claude-opus-4-8 |
| Note | First publicly available Mythos-class model (above Opus); Mythos 5 weights with safety classifiers active. 1M context, $10/$50, highest benchmarks of any GA model. Refusals return stop_reason refusal; mandatory 30-day retention. | Current flagship; 1.6T total / 49B active MoE, 384K max output, optional thinking mode that subsumes R1; 80.6% SWE-bench Verified; MIT license; $0.435/$0.87 per 1M. | Designated current flagship; reliability-focused Opus (4x less likely than 4.7 to let its own code flaws pass). 1M context, $5/$25, effort levels through xhigh/max, Fast mode at $10/$50. Guide positions Fable 5 one tier above it. |
|---|
