BlogLLM Benchmark Scores as Published, Side by Side
DeepSeek-V3.2: Benchmark Scores as Published
Every result the catalog holds for DeepSeek-V3.2: 32 tests in 10 areas, from 7 pages. 8 printed by DeepSeek itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer DeepSeekReleased 2025-12-01Input TextOutput TextWeights Open (MIT)Data version 2026103001, as of 2026-10-30
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Vending-Bench 2 (final balance in USD) | $1034 | Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page | - |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 73.1% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 5 more pages | Effort: thinking |
| SWE-bench Multilingual | 70.2% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 2 more pages | Harness: in-house SWE-agent variant |
| Terminal-Bench 2.0 | 46.4% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 5 more pages | Effort: thinking |
| Terminal-Bench Hard | 35.4% | Z.ai, GLM-4.7 model cardchecked 2026-10-15 | - |
ML Engineering
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| PaperBench | 47.1 | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | - |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 82.4% | Microsoft, Microsoft AI: Introducing MAI-Thinking-1checked 2026-10-19; printed on 4 more pages | - |
| AIME 2025 | 93.1% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 3 more pages | Effort: thinking |
| SciCode | 38.9% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | - |
| HMMT February 2025 | 90.2% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 2 more pages | Effort: thinking |
| IMO-AnswerBench | 78.3% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 3 more pages | - |
| AIME 2026 | 92.7% | Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page | - |
| HMMT November 2025 | 90.2% | Z.ai, GLM-4.7 model cardchecked 2026-10-15; printed on 2 more pages | - |
| HMMT February 2026 | 79.9% | Z.ai, GLM-5.1 model cardchecked 2026-10-15 | - |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LongBench v2 | 59.8% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | Context: 128000 |
Cybersecurity
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| CyberGym | 17.3% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 2 more pages | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Humanity's Last Exam (no tools) | 25.1% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 4 more pages | Effort: thinking |
| MMLU-Pro | 85% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 1 more page | - |
| Humanity's Last Exam (with tools) | 40.8% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 3 more pages | Tools: search, code interpreter, browsing |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| LiveCodeBench v6 | 83.3% | Microsoft, Microsoft AI: Introducing MAI-Thinking-1checked 2026-10-19; printed on 2 more pages | - |
| Codeforces rating | 2386 | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15 | Effort: thinking |
| OJBench (C++) | 54.7% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | - |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Toolathlon | 35.2% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 2 more pages | Effort: thinking |
| MCP Atlas (public set) | 62.2% | Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page | - |
| τ²-bench (domain not stated) | 80.3% | DeepSeek, DeepSeek-V3.2 model cardchecked 2026-10-15; printed on 2 more pages | Effort: thinking |
| τ³-bench (domain not stated) | 69.2% | Z.ai, GLM-5.1 model cardchecked 2026-10-15 | - |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| BrowseComp | 51.4% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15; printed on 3 more pages | Tools: browsing |
| BrowseComp-ZH | 65% | Z.ai, GLM-4.7 model cardchecked 2026-10-15; printed on 1 more page | Tools: browsing |
| Seal-0 | 49.5% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | Tools: search, code interpreter, browsing |
| DeepSearchQA (accuracy) | 60.9% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | Tools: search, code interpreter, browsing |
| WideSearch (item-F1) | 32.5% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | Tools: search, code interpreter, browsing |
| FinSearchComp T2 and T3 | 59.1% | Moonshot AI, Kimi K2.5 model cardchecked 2026-10-15 | Tools: search, code interpreter, browsing |
More from DeepSeek
- DeepSeek-V4.1-Flash-1017
- DeepSeek-V4.1-Flash
- DeepSeek-V4-Flash (Preview)
- DeepSeek-V4-Flash-Base
- DeepSeek-V4-Pro (Preview)
- DeepSeek-V4-Pro-Base
- DeepSeek-V3.2-Base
- DeepSeek-V3.2-Speciale
- DeepSeek-V3.2-Exp
- DeepSeek-V3.1-Terminus
- DeepSeek-V3.1
- DeepSeek-R1-0528
- DeepSeek-R1-0528-Qwen3-8B
- DeepSeek-V3-0324
- DeepSeek-R1
- DeepSeek-R1-Distill-Llama-70B
- DeepSeek-R1-Distill-Llama-8B
- DeepSeek-R1-Distill-Qwen-1.5B
- DeepSeek-R1-Distill-Qwen-14B
- DeepSeek-R1-Distill-Qwen-32B
- DeepSeek-R1-Distill-Qwen-7B
- DeepSeek-V3
- DeepSeek-V4-Flash-0731
- DeepSeek-V4-Pro-0813
- DeepSeek-V4-Vision-Exp
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the 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.