| Benchmark | Score | Δ vs prior | Rank (current) | Configuration | Source |
|---|---|---|---|---|---|
τ²-bench (Telecom) | 78.9% | #7 of 10 | leaderboard | benchlm.ai ↗ | |
Artificial Analysis Intelligence Index | 24.7 | #9 of 11 | leaderboard | benchlm.ai ↗ |
| Version | Released | $ in / out | Terminal-Bench | SWE-bench | GPQA | HLE | ARC-AGI-2 | τ²-bench |
|---|---|---|---|---|---|---|---|---|
| DeepSeek V3 | $0.27 / $1.1 | — | 42% | — | — | — | 22.8% | |
| DeepSeek R1 | $0.55 / $2.19 | — | 49.2% | — | — | — | 36.5% | |
| DeepSeek V3.1 | $0.56 / $1.68 | — | 66% | — | — | — | 34.8% | |
| DeepSeek V3.2 | $0.28 / $0.42 | — | — | — | — | — | 78.9% | |
| DeepSeek V4 Flash | $0.14 / $0.28 | 67% | 79% | 88.1% | 34.8% | 61.4% | — | |
| DeepSeek V4 Pro | $0.435 / $0.87 | 54.7% | 80.6% | 90.1% | 42.7% | 61.3% | 96.2% | |
| DeepSeek V4.1 Flash | $0.14 / $0.28 | 74.5% | — | 90.9% | 36.8% | — | — |