// head_to_head
Qwen3.8 27B vs Step 5 Preview
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.
Step 5 Preview scores higher, Qwen3.8 27B costs less — it depends on your workload.
Step 5 Preview is ahead by 3.9 points overall, and Qwen3.8 27B lists 2.0× cheaper per blended million tokens. Whether 3.9 points is worth that depends on how much a wrong answer costs you. Qwen3.8 27B also leads on measured cost per point of capability, at $1.53 per point.
qwen
Qwen3.8 27B
- Blended / 1M
- $0.730
- Context
- 1M
- Released
- Aug 14, 2026
- Overall score
- 58.1
stepfun
Step 5 Preview
- Blended / 1M
- $1.43
- Context
- 1.0M
- Released
- Oct 8, 2026
- Overall score
- 62.0
Specs and pricing
| Metric | Qwen3.8 27B | Step 5 Preview |
|---|---|---|
| LiveBench overall Mean of the seven LiveBench category scores, 0–100. Higher is better. | 58.1 | 62.0win |
| Cost per point Measured benchmark spend divided by overall score — dollars per point of capability. | $1.53win | $4.08 |
| Blended price / 1M 3:1 input:output mix, the usual shape of production traffic. | $0.730win | $1.43 |
| Input price / 1M | $0.240win | $1.00 |
| Output price / 1M | $2.20win | $2.70 |
| Cached input / 1M Price of an input token served from the prompt cache, where the provider publishes one. | $0.050win | $0.100 |
| Context window | 1M | 1.0M |
| Max output tokens | 236Kwin | 64K |
Benchmarks by category
An overall score averages away the thing you probably care about. These are the seven LiveBench categories scored separately — Qwen3.8 27B on top, Step 5 Preview 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 | Qwen3.8 27B | Step 5 Preview |
|---|---|---|
| Support chatbot 1.2K in / 400 out × 200K requests | $219.92/mo | $391.20/mo |
| RAG assistant 8K in / 600 out × 100K requests | $248.00/mo | $602.00/mo |
| Coding agent 40K in / 4K out × 20K requests | $261.60/mo | $512.00/mo |
| Document extraction 20K in / 1.5K out × 50K requests | $395.50/mo | $1,158/mo |
| Bulk classification 500 in / 20 out × 5M requests | $725.00/mo | $2,320/mo |
Which should you pick?
You are running this at volume
Qwen3.8 27B
Lowest measured cost per point of capability at $1.53 per point — the gap compounds with every request.
Quality matters more than the bill
Step 5 Preview
Highest overall LiveBench score of the two at 62.0.
The workload is coding or agentic work
Qwen3.8 27B
Leads on coding — 86.8 against 85.6.
Qwen3.8 27B vs Step 5 Preview FAQ
Which is better, Qwen3.8 27B or Step 5 Preview?
Step 5 Preview scores higher, Qwen3.8 27B costs less — it depends on your workload. Step 5 Preview is ahead by 3.9 points overall, and Qwen3.8 27B lists 2.0× cheaper per blended million tokens. Whether 3.9 points is worth that depends on how much a wrong answer costs you. Qwen3.8 27B also leads on measured cost per point of capability, at $1.53 per point.
Is Qwen3.8 27B cheaper than Step 5 Preview?
Qwen3.8 27B is cheaper. On a 3:1 input:output blend, Qwen3.8 27B lists at $0.730 per million tokens and Step 5 Preview at $1.43 — Qwen3.8 27B is 2.0× cheaper. Input and output are priced separately — Qwen3.8 27B charges $0.240 in and $2.20 out, Step 5 Preview charges $1.00 and $2.70 — so the model that looks cheaper flips depending on how output-heavy your workload is.
Qwen3.8 27B vs Step 5 Preview: which scores higher on benchmarks?
Qwen3.8 27B scores 58.1 and Step 5 Preview scores 62.0 overall on LiveBench, the mean of its seven categories. That is a 3.9-point lead for Step 5 Preview. Category scores differ from the overall figure — a model can lead on reasoning and trail on coding, which the per-category table above breaks out.
Which gives better value for money, Qwen3.8 27B or Step 5 Preview?
Qwen3.8 27B. Cost per point divides the measured dollars LiveBench spent running the benchmark by the score it earned, so it captures something token pricing misses: a reasoning model can emit many times more tokens than its per-token rate suggests. Qwen3.8 27B works out at $1.53 per point and Step 5 Preview at $4.08.
Does Qwen3.8 27B or Step 5 Preview have a bigger context window?
They are effectively the same — 1M for Qwen3.8 27B and 1.0M for Step 5 Preview.
Do Qwen3.8 27B and Step 5 Preview support prompt caching?
Both publish a cached-input rate: $0.050 per million for Qwen3.8 27B and $0.100 for Step 5 Preview, against full input rates of $0.240 and $1.00. 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 2026-09-02, 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.