// head_to_head
GPT-5.4 Nano vs Qwen3 VL 235B A22B Instruct
Published LiveBench scores across all seven categories, live list pricing, context windows, and the measured cost of a point of capability — for both models, side by side.
GPT-5.4 Nano is the cheaper of the two; neither can be ranked on quality here.
Qwen3 VL 235B A22B Instruct does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
openai
GPT-5.4 Nano
- Blended / 1M
- $0.463
- Context
- 400K
- Released
- Mar 17, 2026
- Overall score
- 69.6
qwen
Qwen3 VL 235B A22B Instruct
- Blended / 1M
- $0.632
- Context
- 262K
- Released
- Sep 23, 2025
- Overall score
- Not evaluated
Specs and pricing
| Metric | GPT-5.4 Nano | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| LiveBench overall Mean of the seven LiveBench category scores, 0–100. Higher is better. | 69.6 | — |
| Cost per point Measured benchmark spend divided by overall score — dollars per point of capability. | $0.0500 | — |
| Blended price / 1M 3:1 input:output mix, the usual shape of production traffic. | $0.463win | $0.632 |
| Input price / 1M | $0.200 | $0.210 |
| Output price / 1M | $1.25win | $1.90 |
| Cached input / 1M Price of an input token served from the prompt cache, where the provider publishes one. | $0.020win | $0.100 |
| Context window | 400Kwin | 262K |
| Max output tokens | 128Kwin | 33K |
Benchmarks by category
An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — GPT-5.4 Nano on top, Qwen3 VL 235B A22B Instruct below, both out of 100.
What each one costs to run
Per-token prices are hard to feel. These are monthly list costs for both models across five workload shapes, using each provider's published cached-input rate where there is one.
| Workload | GPT-5.4 Nano | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $135.04/mo | $194.48/mo |
| RAG assistant 8K in / 600 out × 100K requests | $163.00/mo | $238.00/mo |
| Coding agent 40K in / 4K out × 20K requests | $159.20/mo | $258.40/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $284.75/mo | $347.00/mo |
| Bulk classification 500 in / 20 out × 5M requests | $535.00/mo | $660.00/mo |
Which should you pick?
You need to fit large documents in one call
GPT-5.4 Nano
Wider context window — 400K against 262K.
GPT-5.4 Nano vs Qwen3 VL 235B A22B Instruct FAQ
Which is better, GPT-5.4 Nano or Qwen3 VL 235B A22B Instruct?
GPT-5.4 Nano is the cheaper of the two; neither can be ranked on quality here. Qwen3 VL 235B A22B Instruct does not have a published LiveBench run, so this comparison covers price, context and declared capabilities only. A missing score means "not evaluated", not "worse" — the right way to separate these two is an eval on your own workload.
Is GPT-5.4 Nano cheaper than Qwen3 VL 235B A22B Instruct?
GPT-5.4 Nano is cheaper. On a 3:1 input:output blend, GPT-5.4 Nano lists at $0.463 per million tokens and Qwen3 VL 235B A22B Instruct at $0.632 — GPT-5.4 Nano is 37% cheaper. Input and output are priced separately — GPT-5.4 Nano charges $0.200 in and $1.25 out, Qwen3 VL 235B A22B Instruct charges $0.210 and $1.90 — so the model that looks cheaper flips depending on how output-heavy your workload is.
Does GPT-5.4 Nano or Qwen3 VL 235B A22B Instruct have a bigger context window?
GPT-5.4 Nano has the larger context window: 400K for GPT-5.4 Nano against 262K for Qwen3 VL 235B A22B Instruct. Note that a window you can fill is not a window you should fill — retrieval quality usually degrades well before the limit, and you pay for every token you put in it.
Do GPT-5.4 Nano and Qwen3 VL 235B A22B Instruct support prompt caching?
Both publish a cached-input rate: $0.020 per million for GPT-5.4 Nano and $0.100 for Qwen3 VL 235B A22B Instruct, against full input rates of $0.200 and $0.210. On a workload with a long stable prefix — a system prompt, a tool schema, a retrieved corpus — that changes the economics more than the headline price does.
Related comparisons
How these numbers are produced
- Price — provider list price from OpenRouter, refreshed every 15 minutes. “Blended” is a 3:1 input:output mix.
- Scores — LiveBench release 2026-06-25, using their own category map. Each model shows its strongest published run. A blank means “not evaluated”, never “bad”.
- Cost per point — the measured dollars LiveBench spent on the run, divided by the score it earned.
- “Win” — awarded only past a threshold: one full point on a benchmark score, 10% on a price, 25% on a context window. Anything tighter reports as a tie, because effort settings alone move a LiveBench score by more than that.
Published benchmarks rank models on someone else's tasks. Before committing, see LLM & agent evaluation for building an eval on your own.