The two cheapest million-context models in the catalog.
deepseek-v4-flash · deepseek
gpt-5.6-luna · openai
Benchmark
| Workload Scenario | DeepSeek V4 Flash | GPT-5.6 Luna | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $0.44 | $0.20 | +$0.24 |
| 1M output tokens (generation) | $1.32 | $1.20 | +$0.12 |
| 1M tokens · 70% input / 30% output mix | $0.704 | $0.50 | +$0.204 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.001964 | $0.0014 | +$0.000564 |
| Monthly scale (10K requests / day) | $589.20 | $420.00 | +$169.20 |
Negative difference = DeepSeek V4 Flash is cheaper. Positive = GPT-5.6 Luna is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
GPT-5.6 Luna is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (DeepSeek V4 Flash: 1M, GPT-5.6 Luna: 1.1M) may justify the premium for your task.
FAQ
On input, GPT-5.6 Luna is cheaper ($0.20/M vs $0.44/M). On output, GPT-5.6 Luna is cheaper ($1.20/M vs $1.32/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.001964 on DeepSeek V4 Flash and $0.0014 on GPT-5.6 Luna — GPT-5.6 Luna is 1.4× more expensive for that workload.
DeepSeek V4 Flash supports 1,000,000 tokens (384K max output); GPT-5.6 Luna supports 1,050,000 (128K max output). GPT-5.6 Luna fits 1.1× more context, which matters for long documents and agents.