DeepSeek V4 Pro 0813 vs GPT-6 Astra
GPT-6 Astra leads DeepSeek V4 Pro 0813 on coding, 85.1 to 56.1 on the Vibe Coding Index, and DeepSeek V4 Pro 0813 costs 25.3x less per output token ($1.98 vs $50.00 per million). Updated October 22, 2026; 70 benchmarks have a result for both.
DeepSeek V4 Pro 0813 or GPT-6 Astra: which to pick
Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 56.1 for DeepSeek V4 Pro 0813.
Pick DeepSeek V4 Pro 0813 for output tokens 25.3x cheaper ($1.98 vs $50.00 per million) and work measured by CyberBench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
- Shared results
- 70
- DeepSeek V4 Pro 0813 only
- 8
- GPT-6 Astra only
- 59
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
GPT-6 Astra
GPT-6 Astra scores 85.1 against 56.1.
5 benchmarks behind each score
Intelligence
AA Intelligence Index
GPT-6 Astra
GPT-6 Astra leads, 52.7 against 36.0.
Same board, both measured
Agentic work
AA Agentic Index
GPT-6 Astra
GPT-6 Astra leads, 51.0 against 41.3.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only DeepSeek V4 Pro 0813 has a GPQA Diamond result, 92.4%.
Directional only
Knowledge
MMLU Pro
No call
Only DeepSeek V4 Pro 0813 has a MMLU Pro result, 87.0%.
Directional only
Web design
Design Arena: websites
No call
Only GPT-6 Astra has a Design Arena: websites result, 1336.
Directional only
General
Vals Index
GPT-6 Astra
GPT-6 Astra leads, 65.0% against 53.0%.
Same board, both measured
Long documents
Listed context window
GPT-6 Astra
GPT-6 Astra takes 1.05M tokens against 1.05M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 $0.0016, GPT-6 Astra $0.035 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only DeepSeek V4 Pro 0813 has the number, $0.10.
Directional only
At a glance
- Coding index
- DeepSeek V4 Pro 0813 56.1 from 5 benchmarks, GPT-6 Astra 85.1 from 5 benchmarks.
- Widest category
- Math, 26.9 points apart on average over 6 shared benchmarks.
- Widest single gap
- FrontierMath, tier 4: DeepSeek V4 Pro 0813 26.8% against GPT-6 Astra 97.6%, 70.8 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V4 Pro 0813 $1.98 against GPT-6 Astra $50.00 per million output tokens, 25.3x apart.
- Context window
- DeepSeek V4 Pro 0813 takes 1.05M tokens, GPT-6 Astra 1.05M.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | DeepSeek V4 Pro 0813 | Bars | GPT-6 Astra | Ties |
|---|---|---|---|---|
| Coding | 1 | 17 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 3 | 6 | ||
| Reasoning | 0 | 10 | ||
| Math | 0 | 6 | ||
| Knowledge | 3 | 3 | ||
| General | 9 | 10 | ||
| All shared boards | 16 | 54 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index117
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 93.7% | Only GPT-6 Astra measured |
| LMArena WebDev | 1582 | 1774 | GPT-6 Astra |
| AA Coding Index | 68.8 | 76.9 | GPT-6 Astra |
| LiveCodeBench (Vals AI) | 87.5% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| SWE-bench Verified (Vals AI) | 96.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| Terminal-Bench 4 (Vals AI) | 14.1% | 59.6% | GPT-6 Astra |
| Vibe Code Bench (Vals AI) | 82.3% | 89.6% | GPT-6 Astra |
| Code migration (Vals AI) | 41.5% | 67.7% | GPT-6 Astra |
| ProgramBench (Vals AI) | 0.0% | 5.5% | GPT-6 Astra |
| IOI (Vals AI) | 51.6% | 100.0% | GPT-6 Astra |
| Terminal-Bench 2.1 (Vals AI) | 54.7% | 87.3% | GPT-6 Astra |
| SWE-bench Multimodal (Vals AI) | Not measured | 43.3% | Only GPT-6 Astra measured |
| Vibe Code Bench 1-100 (Vals AI) | 17.5% | 27.6% | GPT-6 Astra |
| SkillsBench (Vals AI) | 53.8% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| CyberBench (Vals AI) | 72.9% | 41.1% | DeepSeek V4 Pro 0813 |
| SciCode | 51.0% | 56.5% | GPT-6 Astra |
| FrontierCode | 28.5% | 53.3% | GPT-6 Astra |
| FrontierSWE | Not measured | 65.5% | Only GPT-6 Astra measured |
| GSO-Bench | Not measured | 79.4% | Only GPT-6 Astra measured |
| WeirdML | 66.2% | 93.3% | GPT-6 Astra |
| MirrorCode | Not measured | 46.7% | Only GPT-6 Astra measured |
| ALE-Bench | 1403.2 | 2951.3 | GPT-6 Astra |
| LMArena Text: coding | 1469 | 1489 | GPT-6 Astra |
| LMArena Text: software and IT | 1462 | 1474 | GPT-6 Astra |
| LMArena WebDev: React | 1588 | 1798 | GPT-6 Astra |
| LMArena WebDev: HTML | 1566 | 1775 | GPT-6 Astra |
| LMArena WebDev: from an image | Not measured | 1726 | Only GPT-6 Astra measured |
| SciCode | Not measured | 12.3% | Only GPT-6 Astra measured |
| SWE-Bench Pro V2 | Not measured | 96.9% | Only GPT-6 Astra measured |
| SWE Atlas: Codebase QnA | Not measured | 59.1% | Only GPT-6 Astra measured |
| SWE Atlas: Refactoring | Not measured | 59.1% | Only GPT-6 Astra measured |
| SWE Atlas: Test Writing | Not measured | 50.7% | Only GPT-6 Astra measured |
| Neither model has a result on the other 11 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| AA Intelligence Index | 36.0 | 52.7 | GPT-6 Astra |
| Epoch Capabilities Index | 155.6 | 168.8 | GPT-6 Astra |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index36
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| AA Agentic Index | 41.3 | 51.0 | GPT-6 Astra |
| Design Arena: agents building web apps | Not measured | 1374 | Only GPT-6 Astra measured |
| Design Arena: agents building full-stack apps | Not measured | 1357 | Only GPT-6 Astra measured |
| Design Arena: agents building mobile apps | Not measured | 1312 | Only GPT-6 Astra measured |
| Design Arena: agents building Godot games | Not measured | 1348 | Only GPT-6 Astra measured |
| Design Arena: PowerPoint slides | Not measured | 1217 | Only GPT-6 Astra measured |
| Design Arena: spreadsheets | Not measured | 1378 | Only GPT-6 Astra measured |
| LiveBench research | Not measured | 85.2% | Only GPT-6 Astra measured |
| LiveBench browser | Not measured | 84.0% | Only GPT-6 Astra measured |
| LiveBench long agentic | Not measured | 82.4% | Only GPT-6 Astra measured |
| SREBench (Vals AI) | 1.9% | 56.9% | GPT-6 Astra |
| Finance Agent (Vals AI) | 50.4% | 53.5% | GPT-6 Astra |
| Tax Agent Bench (Vals AI) | 26.8% | 20.7% | DeepSeek V4 Pro 0813 |
| Legal Research Bench (Vals AI) | 40.9% | 39.4% | DeepSeek V4 Pro 0813 |
| Harvey's Legal Agent Benchmark (Vals AI) | 10.8% | 7.1% | DeepSeek V4 Pro 0813 |
| Excel Modeling Benchmark (Vals AI) | 52.8% | 71.7% | GPT-6 Astra |
| BioMysteryBench (Vals AI) | Not measured | 79.3% | Only GPT-6 Astra measured |
| Terminal-Bench-Science (Vals AI) | 4.3% | 62.9% | GPT-6 Astra |
| Vals Science Index | Not measured | 65.1% | Only GPT-6 Astra measured |
| Time Horizon Index (Vals AI) | Not measured | 90.5% | Only GPT-6 Astra measured |
| CUA-bench (Vals AI) | Not measured | 19.2% | Only GPT-6 Astra measured |
| Robo-bench (Vals AI) | Not measured | 23.3% | Only GPT-6 Astra measured |
| APEX-Agents | 47.3% | 64.7% | GPT-6 Astra |
| Remote Labor Index | Not measured | 20.8% | Only GPT-6 Astra measured |
| BALROG | Not measured | 68.3% | Only GPT-6 Astra measured |
| SpireClimb | Not measured | 25.7% | Only GPT-6 Astra measured |
| Vending-Bench 2 | Not measured | $15,515 | Only GPT-6 Astra measured |
| Neither model has a result on the other 15 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index010
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| GPQA Diamond | 92.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LiveBench reasoning | Not measured | 92.7% | Only GPT-6 Astra measured |
| MysteryMechanism (Vals AI) | Not measured | 53.1% | Only GPT-6 Astra measured |
| GPQA Diamond (Artificial Analysis run) | 92.8% | 96.3% | GPT-6 Astra |
| Humanity's Last Exam (Artificial Analysis run) | 41.0% | 54.7% | GPT-6 Astra |
| GPQA Diamond (Epoch AI run) | 91.7% | 95.8% | GPT-6 Astra |
| ARC-AGI | 90.5% | 98.5% | GPT-6 Astra |
| ARC-AGI-2 | 61.3% | 95.0% | GPT-6 Astra |
| Chess puzzles | 46.0% | 74.0% | GPT-6 Astra |
| Mystery game puzzles | 43.0% | 84.0% | GPT-6 Astra |
| EBR-bench | Not measured | 76.2% | Only GPT-6 Astra measured |
| LMArena Text: hard prompts | 1458 | 1463 | GPT-6 Astra |
| LMArena Text: expert | 1459 | 1489 | GPT-6 Astra |
| LMArena Text: hard prompts in English | 1463 | 1464 | GPT-6 Astra |
| LMArena Vision: diagrams | Not measured | 1297 | Only GPT-6 Astra measured |
| LMArena Vision: homework | Not measured | 1337 | Only GPT-6 Astra measured |
| Humanity's Last Exam (Scale run) | Not measured | 60.6% | Only GPT-6 Astra measured |
| Neither model has a result on the other 6 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| OTIS Mock AIME | 98.6% | 100.0% | GPT-6 Astra |
| FrontierMath, tiers 1 to 3 | 64.6% | 93.7% | GPT-6 Astra |
| FrontierMath, tier 4 | 26.8% | 97.6% | GPT-6 Astra |
| ProofBench | 50.0% | 99.0% | GPT-6 Astra |
| LMArena Text: math | 1441 | 1465 | GPT-6 Astra |
| LMArena Text: mathematical work | 1443 | 1484 | GPT-6 Astra |
| MathArena | Not measured | 81.9% | Only GPT-6 Astra measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index33
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| MMLU Pro | 87.0% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| LegalBench (Vals AI) | 82.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| CorpFin (Vals AI) | 65.4% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| TaxEval (Vals AI) | 73.1% | Not measured | Only DeepSeek V4 Pro 0813 measured |
| MedCode (Vals AI) | 42.5% | 48.5% | GPT-6 Astra |
| SimpleQA Verified | 52.9% | 75.6% | GPT-6 Astra |
| LMArena Text: science | 1457 | 1451 | DeepSeek V4 Pro 0813 |
| LMArena Text: medicine and healthcare | 1446 | 1425 | DeepSeek V4 Pro 0813 |
| LMArena Text: legal and government | 1459 | 1462 | GPT-6 Astra |
| LMArena Text: business and finance | 1442 | 1439 | DeepSeek V4 Pro 0813 |
| FACTS Benchmark Suite | Not measured | 75.7% | Only GPT-6 Astra measured |
| PRBench Finance | Not measured | 47.5% | Only GPT-6 Astra measured |
| PRBench Legal | Not measured | 48.4% | Only GPT-6 Astra measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1336 | Only GPT-6 Astra measured |
| Design Arena: UI components | Not measured | 1394 | Only GPT-6 Astra measured |
| Design Arena: game dev | Not measured | 1480 | Only GPT-6 Astra measured |
| Design Arena: data viz | Not measured | 1334 | Only GPT-6 Astra measured |
| Design Arena: 3D | Not measured | 1500 | Only GPT-6 Astra measured |
| Design Arena: all code categories | Not measured | 1379 | Only GPT-6 Astra measured |
| Design Arena: SVG | Not measured | 1430 | Only GPT-6 Astra measured |
| Design Arena: ASCII art | Not measured | 1353 | Only GPT-6 Astra measured |
| LiveBench design | Not measured | 82.0% | Only GPT-6 Astra measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index910
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| Vals Index | 53.0% | 65.0% | GPT-6 Astra |
| LiveBench data analysis | Not measured | 83.0% | Only GPT-6 Astra measured |
| SAGE (Vals AI) | Not measured | 46.4% | Only GPT-6 Astra measured |
| MedScribe (Vals AI) | 80.2% | 87.9% | GPT-6 Astra |
| Long context reasoning (AA-LCR) | 80.3% | 80.7% | GPT-6 Astra |
| DTBench | 93.9% | 97.3% | GPT-6 Astra |
| LMCA | 45.5% | 64.4% | GPT-6 Astra |
| Furniture assembly | Not measured | 80.0% | Only GPT-6 Astra measured |
| LMArena Text | 1446 | 1441 | DeepSeek V4 Pro 0813 |
| LMArena Text: instruction following | 1448 | 1455 | GPT-6 Astra |
| LMArena Text: creative writing | 1439 | 1420 | DeepSeek V4 Pro 0813 |
| LMArena Vision | Not measured | 1280 | Only GPT-6 Astra measured |
| LMArena Text: English | 1457 | 1447 | DeepSeek V4 Pro 0813 |
| LMArena Text: non-English | 1431 | 1432 | GPT-6 Astra |
| LMArena Text: Chinese | 1481 | 1489 | GPT-6 Astra |
| LMArena Text: German | Not measured | 1459 | Only GPT-6 Astra measured |
| LMArena Text: Japanese | Not measured | 1405 | Only GPT-6 Astra measured |
| LMArena Text: Korean | 1449 | 1428 | DeepSeek V4 Pro 0813 |
| LMArena Text: French | 1446 | 1432 | DeepSeek V4 Pro 0813 |
| LMArena Text: Spanish | 1435 | 1422 | DeepSeek V4 Pro 0813 |
| LMArena Text: Russian | 1448 | 1438 | DeepSeek V4 Pro 0813 |
| LMArena Text: Polish | Not measured | 1422 | Only GPT-6 Astra measured |
| LMArena Text: multi-turn | 1454 | 1447 | DeepSeek V4 Pro 0813 |
| LMArena Text: longer queries | 1458 | 1460 | GPT-6 Astra |
| LMArena Text: writing and language | 1443 | 1437 | DeepSeek V4 Pro 0813 |
| LMArena Text: entertainment and media | 1424 | 1427 | GPT-6 Astra |
| LMArena Vision: English | Not measured | 1280 | Only GPT-6 Astra measured |
| LMArena Vision: Chinese | Not measured | 1361 | Only GPT-6 Astra measured |
| LMArena Vision: reading text | Not measured | 1292 | Only GPT-6 Astra measured |
| LMArena Vision: creative writing | Not measured | 1286 | Only GPT-6 Astra measured |
| Neither model has a result on the other 18 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V4 Pro 0813 | GPT-6 Astra | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
DeepSeek V4 Pro 0813 costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Pro 0813 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Pro 0813 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | DeepSeek V4 Pro 0813 | OpenAI GPT-6 Astra |
|---|---|---|
| Maker | DeepSeek | OpenAI |
| Weights | Open weights | No public weights listed |
| Released | August 12, 2026 | September 4, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.66 | $10.00 |
| Output $/Mtok | $1.98 | $50.00 |
| Coding index | 56.1 | 85.1 |
| Independent measurements behind it | 5 | 5 |
| Cost per solved task | $0.10 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V4 Pro 0813 better than GPT-6 Astra for coding?
Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 56.1 for DeepSeek V4 Pro 0813. Pick DeepSeek V4 Pro 0813 for output tokens 25.3x cheaper ($1.98 vs $50.00 per million) and work measured by CyberBench (Vals AI) or Tax Agent Bench (Vals AI), where it leads.
Which is cheaper, DeepSeek V4 Pro 0813 or GPT-6 Astra?
DeepSeek V4 Pro 0813 costs $0.66 per million input tokens and $1.98 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. Cost per solved task is known for DeepSeek V4 Pro 0813 only.
Which has the bigger context window, DeepSeek V4 Pro 0813 or GPT-6 Astra?
GPT-6 Astra, with 1.05M tokens against 1.05M.
How many benchmarks is the DeepSeek V4 Pro 0813 vs GPT-6 Astra comparison based on?
70 benchmarks have a result for both models as of October 22, 2026. DeepSeek V4 Pro 0813's coding index rests on 5 independent benchmarks and GPT-6 Astra's on 5. Where only one model has a result, the page says so instead of naming a winner.