head-to-head
| Metric | Claude Opus 4.8 | DeepSeek V4 Pro |
|---|---|---|
| SWE-bench Verified | ~86% | 80.6% |
| SWE-bench Pro | 69.2% | 55.4% |
| Terminal-Bench | ~82.7% (TB2.1) | 67.9% (TB2.0) |
| Input $ / 1M | $5 | $0.435 |
| Context | 1M | 1M |
| Open weights | No | Yes |
| Maker | Anthropic | DeepSeek |
when to pick each
Pick Claude Opus 4.8 if
The hardest agentic refactors and long, autonomous multi-file tasks where every point of accuracy saves a human review cycle.
Pick DeepSeek V4 Pro if
The cheapest frontier-class coder — top open-weights score at ~11× less than Opus. Best pick when cost or self-hosting rules.
Ranked on our AI Coding Leaderboard, updated 2026-07-02. Scores are confirmed against primary sources; prices are per 1M input tokens and can change.
Primary sources
- AnthropicAnthropic — Claude Opus 4.8 — Anthropic-reported; independent evals (vals.ai) track within ~1 point.
- DeepSeekDeepSeek V4 — specs & benchmarks — Independent tracker (llm-stats, June 2026); tied with Gemini 3.1 Pro on Verified, ahead on Pro.
- BenchmarkSWE-bench — the real-GitHub-issue benchmark