DeepSeek V4 Pro vs GPT-5.6 Luna
| Spec | DeepSeek V4 Pro | GPT-5.6 Luna |
|---|---|---|
| Respan Index | 49.4 (#35) | 48.8 (#37) |
| Blended price per 1M | $0.427 | $0.225 |
| Context window | 1M | 1.1M |
| Input / output per 1M | $0.28 / $0.87 | $0.1 / $0.6 |
| Max output | 384K | 128K |
| Input | text | text, vision |
| Released | 2026-08-13 | 2026-07-09 |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | DeepSeek V4 Pro | GPT-5.6 Luna |
|---|---|---|
| Coding | 30.6 DeepSeek V4 Pro | 27.8 GPT-5.6 Luna |
| Agents | 59.3 DeepSeek V4 Pro | 56.9 GPT-5.6 Luna |
| Reasoning | 53.4 DeepSeek V4 Pro | 62.6 GPT-5.6 Luna |
| Knowledge & language | 52.6 DeepSeek V4 Pro | 38.7 GPT-5.6 Luna |
| Human preference | 55.5 DeepSeek V4 Pro | 53.8 GPT-5.6 Luna |
Benchmarks both report
DeepSeek V4 Pro leads on 17 of the 25 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | DeepSeek V4 Pro | GPT-5.6 Luna |
|---|---|---|
| Chess Puzzles (Epoch AI) | 46% | 43% |
| LMArena Agent (LMArena) | -0.003 | -0.0081 |
| LMArena Elo | 1465 | 1453 |
| LMArena WebDev (LMArena) | 1582 | 1519 |
| APEX-Agents (Mercor) | 47.3% | 43% |
| DeepSWE v1.1 (Datacurve) | 62.8% | 67.2% |
| FrontierMath Tier 4 v2 (Epoch AI) | 26.8% | 61% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 64.6% | 82.1% |
| GPQA Diamond | 91.7% | 92.3% |
| LiveBench Agentic Coding (LiveBench) | 54.9% | 48.4% |
| LiveBench Coding (LiveBench) | 77.2% | 82.9% |
| LiveBench Data Analysis (LiveBench) | 79.2% | 78% |
| LiveBench Instruction Following (LiveBench) | 67.7% | 60.1% |
| LiveBench Language (LiveBench) | 82.1% | 72.6% |
| LiveBench Mathematics (LiveBench) | 95.1% | 87.2% |
| LiveBench Reasoning (LiveBench) | 85.8% | 85.6% |
| Mystery Game Puzzles (Epoch AI) | 43% | 21% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.6% | 98.3% |
| SimpleQA Verified | 52.9% | 41% |
| SWE-rebench 2026-05-15 to 2026-07-01 (Nebius) | 40.2% | 43.6% |
| Terminal-Bench 4.0 (Vals AI) | 14.1% | 11.6% |
| Vending-Bench 2 (Andon Labs) | 3284.52 | 4094.71 |
| WeirdML (Håvard Tveit Ihle) | 66.2% | 60.9% |
| DeepSWE v1.1 | 62.7% | 67.2% |
| SimpleBench (SimpleBench) | 50.9% | 46.8% |
FAQ
Is DeepSeek V4 Pro better than GPT-5.6 Luna?
DeepSeek V4 Pro scores 49.4 on the Respan Index (#35), ahead of GPT-5.6 Luna at 48.8 (#37). DeepSeek V4 Pro leads on 17 of the 25 benchmarks both report.
Which is cheaper, DeepSeek V4 Pro or GPT-5.6 Luna?
DeepSeek V4 Pro lists at $0.28 input and $0.87 output per 1M tokens; GPT-5.6 Luna at $0.1 and $0.6. GPT-5.6 Luna is 1.9x cheaper per token at list price.
Which has the larger context window?
DeepSeek V4 Pro accepts 1M tokens and GPT-5.6 Luna 1.1M tokens.