BlogLLM Benchmark Scores as Published, Side by Side
GPT-6 Astra vs DeepSeek-V4.1-Flash-1017: Benchmark Scores Side by Side
Every test both GPT-6 Astra by OpenAI and DeepSeek-V4.1-Flash-1017 by DeepSeek have a published number for: 43 tests in 9 areas. Each number is as printed, with the page it comes from and the setup it was run under; nothing is ranked or rescored, and two numbers under different setups are two results, not one comparison.
Models GPT-6 Astra, DeepSeek-V4.1-Flash-1017Data version 2026103001, as of 2026-10-30
Model
What each developer prints on its own model page; a dash means it prints none.
| Fact | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| Released | 2026-09-03 | 2026-10-17 |
| Context window | 1.05M | 1M |
| Max output | 128K | - |
| Input | Text, image | Text, image |
| Output | Text | Text |
| Weights | Closed | Open (MIT) |
Knowledge Work
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| AutomationBench | 41.4%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 7 more pages | 60.5%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 2 more pagesDataset revision: v1.0.6 (printed in the label) |
| APEX-Agents | ||
| OfficeQA Pro | 66.2%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Codex; Fallback: with fallback | |
| GDPval-AA v2.1 | 1542Anthropic, Claude Opus 5.5, checked 2026-10-14; printed on 3 more pages | |
| AA-Briefcase v1.1 | 1569Anthropic, Claude Opus 5.5 System Card, checked 2026-10-25; printed on 1 more page | |
| GDP.pdf w/ python (mean rubric pass rate) | 85.5%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Fallback: with fallback | |
| Business Operation (Internal) | ||
| FrontierFinance | 45%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Finance Agent v2; Fallback: with fallback | 54.5%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Finance Agent v2 |
| OfficeQA Pro v2 | 54.4%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Codex; Fallback: with fallback | |
| DECK-Arena (Internal, Elo) |
Agentic Coding
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| DeepSWE v1.1 | 74.1%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 3 more pages | 77%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 2 more pagesHarness: Mini-SWE-V2 |
| Terminal-Bench 4.0 | 57.9%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 7 more pages | 42.3%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 2 more pagesHarness: Claude Code |
| ProgramBench Almost@1 | 52.5%Meta, Meta Muse Wonder model card, checked 2026-10-19; printed on 1 more pageEffort: max; Harness: mini-swe-agent | 29.1%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 1 more page |
| SWE-Marathon | 40%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Claude Code; Dataset revision: v1.1 | |
| Kimi Code Bench 3.0 Long-horizon |
ML Engineering
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| PostTrainBench v1.2 | 41.9%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Dataset revision: v1.2; Fallback: with fallback | 30.5%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Claude Code; Dataset revision: v1.2 |
Science and Math
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| GPQA Diamond | 96%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 1 more page | |
| Terminal-Bench Science 0.1 | 64.6%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 7 more pages | 9.3%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Model as printed: DeepSeek-V4.1-Flash-1017 15.7%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Model as printed: DeepSeek-V4.1-Flash-1017 (max) |
| FrontierMath Tier 4 (v2) | 97.6%OpenAI, GPT-6 Astra, checked 2026-10-14 | |
| FrontierMath Tier 1-3 (v2) | ||
| BioMysteryBench | 78.9%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Codex; Fallback: with fallback |
Computer Use
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| OSWorld 2.1 partial |
Multimodal Understanding
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| CharXiv Reasoning (no tools) | 90%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | |
| MMMU-Pro | 86.9%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | |
| GDP.pdf | ||
| CharXiv Reasoning (with tools) | 92.8%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Fallback: with fallback | |
| BabyVision (with tools) | 96.6%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Fallback: with fallback | 92.5%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 2 more pagesTools: enabled; exact tools in source |
| MMMU-Pro (with tools) | 87.8%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Fallback: with fallback | |
| Chartography (with tools) | 91%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Fallback: with fallback | 83%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 1 more pageTools: enabled; exact tools in source |
| ZeroBench (with tools, pass@5) | 58%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: python; Attempt budget: pass@5; Fallback: with fallback | 51%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 1 more pageTools: enabled; exact tools in source |
| PerceptionBench | 77.6%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | |
| ZeroBench (pass@5) | 52%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | |
| ZeroBench_main (avg@5) | 43.2%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback |
General Knowledge and Reasoning
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| Humanity's Last Exam (no tools) | 53.6%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | 42.6%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 2 more pages 44%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18Dataset revision: text-only subset of HLE (dagger on the card) |
| Humanity's Last Exam (with tools) | 57.2%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 3 more pagesTools: enabled; exact tools in source | |
| SimpleQA Verified | ||
| Humanity's Last Exam Diamond (no tools) | 65.8%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback | |
| Kimi Hallu Bench Hard w/ tools (lower is better)lower is better | 5.5%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: with tools; Fallback: with fallback | |
| Kimi Hallu Bench w/o tools (lower is better)lower is better | 4.2%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Tools: none; Fallback: with fallback |
Tool Use and Workflows
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| Agents' Last Exam (pass rate) | 34.2%Google, Gemini 4 Argon, checked 2026-10-29; printed on 1 more page | 35.4%DeepSeek, DeepSeek-V4.1-Flash-1017 model card, checked 2026-10-18; printed on 1 more page |
| Agents' Last Exam (CLI) | 38.8%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Dataset revision: v1.1; Fallback: with fallback | 36%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Claude Code; Dataset revision: v1.1 |
Engineering and Design
| Test | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| BenchCAD | 95.9%OpenAI, GPT-6 Astra, checked 2026-10-14; printed on 1 more page | 81.6%Moonshot AI, Introducing Kimi K3.1 Pro, checked 2026-10-24Effort: max; Harness: Claude Code; Tools: with tools |
| Design Bench (Internal, Elo) |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked; a dash means the developer lists that price in another unit or not at all. Cost per task is measured by Vals AI.
| Price | GPT-6 Astra | DeepSeek-V4.1-Flash-1017 |
|---|---|---|
| Input | $10OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $20 per 1M tokens above 272,000 tokens | $0.3DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Cached input | $1OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $2 per 1M tokens above 272,000 tokens | $0.006DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Cache write | $12.5OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $25 per 1M tokens above 272,000 tokens | - |
| Output | $50OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $75 per 1M tokens above 272,000 tokens | $1.2DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Input, batch | $5OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $10 per 1M tokens above 272,000 tokens | - |
| Cached input, batch | $0.5OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $1 per 1M tokens above 272,000 tokens | - |
| Cache write, batch | $6.25OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $12.5 per 1M tokens above 272,000 tokens | - |
| Output, batch | $25OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $37.5 per 1M tokens above 272,000 tokens | - |
| Input, flex | $5OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $10 per 1M tokens above 272,000 tokens | - |
| Cached input, flex | $0.5OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $1 per 1M tokens above 272,000 tokens | - |
| Cache write, flex | $6.25OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $12.5 per 1M tokens above 272,000 tokens | - |
| Output, flex | $25OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $37.5 per 1M tokens above 272,000 tokens | - |
| Input, fast | $20OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $40 per 1M tokens above 272,000 tokens | - |
| Cached input, fast | $2OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $4 per 1M tokens above 272,000 tokens | - |
| Cache write, fast | $25OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $50 per 1M tokens above 272,000 tokens | - |
| Output, fast | $100OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $150 per 1M tokens above 272,000 tokens | - |
| Input, ultrafast | $60OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $120 per 1M tokens above 272,000 tokens | - |
| Cached input, ultrafast | $6OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $12 per 1M tokens above 272,000 tokens | - |
| Cache write, ultrafast | $75OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $150 per 1M tokens above 272,000 tokens | - |
| Output, ultrafast | $300OpenAI, OpenAI API pricing, checked 2026-10-15Context: up to 272,000 tokens; Long context: $450 per 1M tokens above 272,000 tokens | - |
| Input, off-peak hours | - | $0.15DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Cached input, off-peak hours | - | $0.003DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Output, off-peak hours | - | $0.6DeepSeek, DeepSeek API pricing, checked 2026-10-18 |
| Cost per taskVals AI, USD, lower is better |
GPT-6 Astra has 159 more and DeepSeek-V4.1-Flash-1017 has 7 more: tests only one of them has a number for, listed on its model page.
More side by side
- GPT-6 Astra vs Claude Fable 5.1
- GPT-6 Astra vs Gemini 4 Carbon
- GPT-6 Astra vs Grok 4.7
- GPT-6 Astra vs Qwen3.8-Max
- GPT-6 Astra vs Kimi K3.1 Pro
- GPT-6 Astra vs GLM-5.3
- Claude Fable 5.1 vs DeepSeek-V4.1-Flash-1017
- DeepSeek-V4.1-Flash-1017 vs Qwen3.8-Max
- DeepSeek-V4.1-Flash-1017 vs Kimi K3.1 Pro
- DeepSeek-V4.1-Flash-1017 vs GLM-5.3
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. A pair has a page when both models are their developers' flagships and both have a number for at least 10 of the same tests. 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.