GPT-5.6 Luna vs DeepSeek-V4 Pro
Head-to-head technical comparison of intelligence, capabilities, and API pricing.
Which should you use?
GPT-5.6 Luna and DeepSeek-V4 Pro score within 0.2 points of each other on the composite index, which is too close to call a winner on capability alone.
Their Arena Elo differs by only 6 points, inside the confidence intervals the public board reports, so human preference does not separate them.
Cost is the sharper difference: DeepSeek-V4 Pro is 2.2x the input price of GPT-5.6 Luna ($0.435 against $0.2 per 1M), so at volume the choice is usually decided by budget rather than benchmarks.
Pick GPT-5.6 Luna when
- Stronger on MMLU-Pro: 92% against 91.8%.
- Cheaper input tokens — $0.2 against $0.435 per 1M, so DeepSeek-V4 Pro costs 2.2x as much to feed.
- Better value per point of measured capability: $0.002 per index point against $0.005.
Pick DeepSeek-V4 Pro when
On these measures DeepSeek-V4 Pro does not lead GPT-5.6 Luna anywhere, so pick it only for reasons outside this table — an existing integration, a region, or a contract.
Value per point of measured capability, at list input pricing: GPT-5.6 Luna $0.002 · DeepSeek-V4 Pro $0.005 per index point. Run your own token mix through the token cost calculator — a comparison at list price ignores caching and batch discounts, which move real bills more than this gap does.
Where these numbers come from
Figures as of 2026-09-14. Arena Elo and pricing are read from the public board and each vendor's own pricing page; see the full leaderboard for all 34 models and which columns are measured rather than estimated, or the benchmark matrix for scores benchmark by benchmark.
Our reporting on OpenAI and DeepSeek
OpenAI EU Text Watermark: 25% Word Swaps Drop Detection to 17%
OpenAI is rolling out invisible textGrain watermarks to EU ChatGPT and Codex output, but its own tests show replacing 25% of words drops detection from 92% to 17%.
12 Lines of Code Could Hijack ChatGPT’s Mac App
A patched ChatGPT macOS flaw let local malicious code reach chat logs and issue trusted commands through the app, WIRED reports.
Generate and Run a Python Expression with Outlines
Use Outlines and GPT-4o mini to turn a word problem into a Python expression, execute it, and understand the risks of eval.