Open-weight frontier versus the OpenAI default tier — with off-peak twists.
deepseek-v4-pro · deepseek
gpt-5.6-terra · openai
Benchmark
| Workload Scenario | DeepSeek V4 Pro | GPT-5.6 Terra | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $1.32 | $2.00 | −$0.68 |
| 1M output tokens (generation) | $3.96 | $12.00 | −$8.04 |
| 1M tokens · 70% input / 30% output mix | $2.112 | $5.00 | −$2.888 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.005896 | $0.014 | −$0.008104 |
| Monthly scale (10K requests / day) | $1,768.80 | $4,200.00 | −$2,431.20 |
Negative difference = DeepSeek V4 Pro is cheaper. Positive = GPT-5.6 Terra is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
DeepSeek V4 Pro 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 Pro: 1M, GPT-5.6 Terra: 1.1M) may justify the premium for your task.
FAQ
On input, DeepSeek V4 Pro is cheaper ($1.32/M vs $2.00/M). On output, DeepSeek V4 Pro is cheaper ($3.96/M vs $12.00/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 $0.005896 on DeepSeek V4 Pro and $0.014 on GPT-5.6 Terra — DeepSeek V4 Pro is 2.4× cheaper for that workload.
DeepSeek V4 Pro supports 1,000,000 tokens (384K max output); GPT-5.6 Terra supports 1,050,000 (128K max output). GPT-5.6 Terra fits 1.1× more context, which matters for long documents and agents.