DeepSeek V4 Pro vs Gemini 3.8 Flash
| Spec | DeepSeek V4 Pro | Gemini 3.8 Flash |
|---|---|---|
| Respan Index | 49.2 (#45) | 54.5 (#31) |
| Blended price per 1M | $0.585 | $1.5 |
| Context window | 1M | 1.0M |
| Input / output per 1M | $0.28 / $1.5 | $0.75 / $3.75 |
| Max output | 384K | 66K |
| Input | text | text, vision, audio |
| Released | 2026-08-13 | 2026-09-02 |
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 | Gemini 3.8 Flash |
|---|---|---|
| Coding | 29.8 DeepSeek V4 Pro | 36.1 Gemini 3.8 Flash |
| Agents | 59.7 DeepSeek V4 Pro | 66.7 Gemini 3.8 Flash |
| Reasoning | 52.8 DeepSeek V4 Pro | 53.4 Gemini 3.8 Flash |
| Knowledge & language | 52.6 DeepSeek V4 Pro | 70.9 Gemini 3.8 Flash |
| Human preference | 55.6 DeepSeek V4 Pro | 56.0 Gemini 3.8 Flash |
Benchmarks both report
Gemini 3.8 Flash leads on 20 of the 25 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | DeepSeek V4 Pro | Gemini 3.8 Flash |
|---|---|---|
| LMArena WebDev (LMArena) | 1581 | 1584 |
| LMArena Elo | 1464 | 1497 |
| LMArena Agent (LMArena) | -0.003 | 0.0149 |
| MirrorCode (Epoch AI) | 33.3% | 38.9% |
| Chess Puzzles (Epoch AI) | 46% | 58% |
| APEX-Agents (Mercor) | 47.3% | 64.3% |
| DeepSWE v1.1 (Datacurve) | 62.8% | 73.8% |
| FrontierMath Tier 4 v2 (Epoch AI) | 26.8% | 22% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 64.6% | 68.4% |
| GPQA Diamond | 91.7% | 95.4% |
| LiveBench Agentic Coding (LiveBench) | 54.9% | 54.2% |
| LiveBench Coding (LiveBench) | 77.2% | 72.5% |
| LiveBench Data Analysis (LiveBench) | 79.2% | 54% |
| LiveBench Instruction Following (LiveBench) | 67.7% | 81.4% |
| LiveBench Language (LiveBench) | 82.1% | 87.8% |
| LiveBench Mathematics (LiveBench) | 95.1% | 91.6% |
| LiveBench Reasoning (LiveBench) | 85.8% | 89.3% |
| Mystery Game Puzzles (Epoch AI) | 43% | 47% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.6% | 98.9% |
| SimpleQA Verified | 52.9% | 69.7% |
| Terminal-Bench 4.0 (Vals AI) | 14.1% | 19.7% |
| Vending-Bench 2 (Andon Labs) | 3284.52 | 5093.79 |
| WeirdML (Håvard Tveit Ihle) | 66.2% | 84.8% |
| DeepSWE v1.1 | 62.7% | 73.7% |
| SimpleBench (SimpleBench) | 50.9% | 82.4% |
FAQ
Is DeepSeek V4 Pro better than Gemini 3.8 Flash?
Gemini 3.8 Flash scores 54.5 on the Respan Index (#31), ahead of DeepSeek V4 Pro at 49.2 (#45). Gemini 3.8 Flash leads on 20 of the 25 benchmarks both report.
Which is cheaper, DeepSeek V4 Pro or Gemini 3.8 Flash?
DeepSeek V4 Pro lists at $0.28 input and $1.5 output per 1M tokens; Gemini 3.8 Flash at $0.75 and $3.75. DeepSeek V4 Pro is 2.6x cheaper per token at list price.
Which has the larger context window?
DeepSeek V4 Pro accepts 1M tokens and Gemini 3.8 Flash 1.0M tokens.
Compare other models
- vsGemini 3.8 Flash vs Gemini 3.7 Flash
- vsClaude Sonnet 5.5 vs Gemini 3.8 Flash
- vsGemini 3.8 Flash vs Claude Sonnet 5
- vsDeepSeek V4 Pro vs GPT-5.4
- vsDeepSeek V4 Pro vs GPT-6 Luna
- vsDeepSeek V4 Pro vs Claude Opus 4.7
- vsGemini 3.8 Flash vs GPT-5.5
- vsGemini 3.8 Flash vs Grok 4.6
- vsGemini 3.8 Flash vs Step 5 Preview