BlogLLM Benchmark Scores as Published, Side by Side
DSBench (DeepSeek internal): Results as Published
Every result the catalog holds for DSBench (DeepSeek internal): 8 models from 4 developers, read from 2 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Agentic CodingMaintainer DeepSeek (internal)Introduced 2026Data version 2026103001, as of 2026-10-30
DSBench is a pair of internal DeepSeek test sets: DSBench-FullStack holds full-stack development tasks and DSBench-Hard holds difficult coding-agent problems. The tasks and the scoring are not published.
DSBench-FullStack
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Fable 5Anthropic, released 2026-06-09 | 77.2% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 71.6% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Flash (Preview)DeepSeek, released 2026-04-24 | 37% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Pro (Preview)DeepSeek, released 2026-04-24 | 41.8% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Flash-0731DeepSeek | 68.7% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Pro-0813DeepSeek | 71.1% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| Kimi K3Moonshot AI, released 2026-07-16 | 73.7% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| GLM-5.2Z.ai, released 2026-06-16 | 61.8% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
DSBench-Hard
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Fable 5Anthropic, released 2026-06-09 | 68.3% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| Claude Opus 4.8Anthropic, released 2026-05-28 | 71.7% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Flash (Preview)DeepSeek, released 2026-04-24 | 25.8% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Pro (Preview)DeepSeek, released 2026-04-24 | 31.1% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Flash-0731DeepSeek | 59.6% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
| DeepSeek-V4-Pro-0813DeepSeek | 67.2% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| Kimi K3Moonshot AI, released 2026-07-16 | 63% | DeepSeek, DeepSeek-V4-Pro-0813 model cardchecked 2026-10-15 | - |
| GLM-5.2Z.ai, released 2026-06-16 | 54.5% | DeepSeek, DeepSeek-V4-Flash-0731 model cardchecked 2026-10-15; printed on 1 more page | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.