DeepSeek V4 vs GPT-5 mini

Benchmarks, API pricing and specs, head to head. Data updated 2026-06-10.

DeepSeek V4

DeepSeek · Mar 2026

The open-weights shock of 2026: ~81% SWE-bench Verified, a 1M context and MIT license at $0.30/$0.50 per million tokens.

GPT-5 mini

OpenAI · Aug 2025

A fast, cheap distillation of GPT-5 for high-volume workloads — one fifth of the flagship's price with most of its everyday capability.

The verdict

DeepSeek V4 wins 1 of the 1 benchmarks these models share, against 0 for GPT-5 mini. DeepSeek V4 is about 2.0x cheaper per blended million tokens (3:1 input:output mix). DeepSeek V4 also takes 1M of context versus 400K for GPT-5 mini. And DeepSeek V4 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.

Benchmark head-to-head 10

SWE-bench Verified
~81%~71%
DeepSeek V4GPT-5 mini

Specs & pricing

DeepSeek V4GPT-5 mini
modhub Index
Input price / 1M$0.3$0.25
Output price / 1M$0.5$2
Context window1M400K
Max output64K128K
Open weightsyes (MIT)no
Reasoning modelyesyes
Multimodal inputtexttext, image
Knowledge cutoffDec 2025May 2024
ReleasedMar 2026Aug 2025
Example monthly cost*$3.75$5.50

* 10M input + 1.5M output tokens per month at list prices, no caching. Green = better value on that row.

Frequently asked questions

Which is better, DeepSeek V4 or GPT-5 mini?
DeepSeek V4 wins 1 of the 1 benchmarks these models share, against 0 for GPT-5 mini. DeepSeek V4 is about 2.0x cheaper per blended million tokens (3:1 input:output mix). DeepSeek V4 also takes 1M of context versus 400K for GPT-5 mini. And DeepSeek V4 is open-weights (MIT), so it can be self-hosted — a structural advantage if data control or vendor independence matters.
Which is cheaper, DeepSeek V4 or GPT-5 mini?
DeepSeek V4 costs $0.3/$0.5 per million input/output tokens, while GPT-5 mini costs $0.25/$2. For a typical workload of 10M input and 1.5M output tokens per month, that's $3.75 versus $5.50.
Which model is better for coding, DeepSeek V4 or GPT-5 mini?
On SWE-bench Verified — the standard agentic-coding benchmark — DeepSeek V4 scores ~81% versus ~71% for GPT-5 mini, making DeepSeek V4 the stronger pick for coding agents.

More comparisons