Based on Z.ai's official comparison table: GPT-5.6 Sol still ranks first on Terminal-Bench 3.0, HLE w/ Tools, and other benchmarks, while V4-Pro holds its place in the open-source top tier at roughly 1/10 the price.
| Benchmark | GLM-5.3 | Kimi K3 | DeepSeek-V4-Pro | Opus 4.8 | GPT-5.6 Sol |
|---|---|---|---|---|---|
| Terminal-Bench 2.1 | 88.2 | 88.3 | 87.9 | 85.0 | 88.8 |
| Terminal-Bench 3.0 | 28.3 | 17.4 | — | 21.1 | 34.6 |
| DeepSWE v1.1 | 66.9 | 67.5 | 62.7 | 58.0 | 72.7 |
| NL2Repo | 58.0 | 58.0 | 61.1 | 69.7 | — |
| CyberGym | 84.5 | 80.0 | 83.3 | 78.1 | 83.6 |
| Toolathlon Verified | 73.0 | 76.5 | 74.1 | 76.2 | 74.9 |
| AutomationBench v1.0.6 | 48.2 | 46.7 | 43.2 | 41.0 | 45.8 |
| ALE-CLI | 28.5 | 27.6 | 25.7 | 25.7 | 28.6 |
| HLE w/ Tools | 62.5 | 59.8 | 60.0 | 57.9 | 64.5 |
| GDPval-AA v2 | 1769 | 1682 | 1590 | 1588 | 1730 |
Data source: Z.ai official blog (2026-08-14); not evaluated is indicated by —. DeepSeek pricing: V4-Pro off-peak $0.66/$1.98, V4-Flash $0.14/$0.28, V4.1-Flash-1017 $0.15/$0.60 per million tokens.
GPT-5.6 Sol leads across the board with TB 2.1 88.8, TB 3.0 34.6, and HLE w/ Tools 64.5; V4-Pro follows closely with 87.9 / not evaluated / 60.0, while its NL2Repo score of 61.1 overtakes Opus 4.8.
V4-Pro off-peak $0.66/$1.98, V4-Flash $0.14/$0.28, cache hit $0.003; closed-source flagships typically cost 5~10x more than open-source at official list prices (refer to each vendor's official website).
V4's weights are MIT-licensed and self-hostable, so your data never leaves the intranet; the GPT series is API-only. For enterprise compliance and offline scenarios, DeepSeek is the default solution.
If you want the absolute best, go with GPT-5.6; if you want 90% of the capability at 1/10 the price plus self-hosting, go with DeepSeek—and adopt a two-tier routing setup with deepseek-flash as the everyday default, escalating hard tasks to V4-Pro.
大模型网摘