Head-to-head showdown: Qwen3.8 27B (OpenRouter) ($0.42 in / $3.00 out per 1M) vs GPT-5.4 Workhorse ($2.50 in / $15.00 out per 1M). Qwen3.8 27B (OpenRouter) is 5.1× cheaper across standard token mixes, with 1M vs 256K context windows.
qwen3.8-27b · qwen
gpt-5.4 · openai
Benchmarks
Qwen3.8 27B (OpenRouter)
No verified third-party composite score published.
GPT-5.4 Workhorse
Third-party composite indices via OpenRouter's live catalog — source and verification date on each model page.
Cost by Volume
| Workload Scenario | Qwen3.8 27B (OpenRouter) | GPT-5.4 Workhorse | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $0.42 | $2.50 | −$2.08 |
| 1M output tokens (generation) | $3.00 | $15.00 | −$12.00 |
| 1M tokens · 70% input / 30% output mix | $1.194 | $6.25 | −$5.056 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.00408 | $0.0175 | −$0.0134 |
| Monthly scale (10K requests / day) | $1,224.00 | $5,250.00 | −$4,026.00 |
Negative difference = Qwen3.8 27B (OpenRouter) is cheaper. Positive = GPT-5.4 Workhorse is cheaper.
Best For
Fit score = weighted composite blend for each use case; value = fit ÷ cost per 1,000 requests. Scoreboard: Qwen3.8 27B (OpenRouter) 0 – 8 GPT-5.4 Workhorse.
| Use case | Fit — Qwen3.8 27B (OpenRouter) | Fit — GPT-5.4 Workhorse | Cost / 1K — Qwen3.8 27B (OpenRouter) | Cost / 1K — GPT-5.4 Workhorse | Winner |
|---|---|---|---|---|---|
| Customer support chatbot | — | 51.3 | $2.52 | $8.488 | GPT-5.4 Workhorse — only scored model |
| RAG / search-augmented answers | — | 54.0 | $4.86 | $18.50 | GPT-5.4 Workhorse — only scored model |
| AI coding assistant | — | 63.0 | $11.04 | $43.80 | GPT-5.4 Workhorse — only scored model |
| Document summarization | — | 52.2 | $12.30 | $65.875 | GPT-5.4 Workhorse — only scored model |
| Agentic workflow | — | 51.4 | $21.30 | $64.00 | GPT-5.4 Workhorse — only scored model |
| Content generation | — | 52.2 | $3.936 | $19.46 | GPT-5.4 Workhorse — only scored model |
| Data extraction & tagging | — | 58.5 | $1.59 | $7.85 | GPT-5.4 Workhorse — only scored model |
| Translation | — | 53.1 | $18.60 | $93.312 | GPT-5.4 Workhorse — only scored model |
Costs use each use case's typical token mix and cacheable share. Fit weights are documented on the methodology page.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
Qwen3.8 27B (OpenRouter) 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 (Qwen3.8 27B (OpenRouter): 1M, GPT-5.4 Workhorse: 256K) may justify the premium for your task.
FAQ
On input, Qwen3.8 27B (OpenRouter) is cheaper ($0.42/M vs $2.50/M). On output, Qwen3.8 27B (OpenRouter) is cheaper ($3.00/M vs $15.00/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 $0.00408 on Qwen3.8 27B (OpenRouter) and $0.0175 on GPT-5.4 Workhorse — Qwen3.8 27B (OpenRouter) is 4.3× cheaper for that workload.
Qwen3.8 27B (OpenRouter) supports 1,000,000 tokens (131.1K max output); GPT-5.4 Workhorse supports 256,000 (32.8K max output). Qwen3.8 27B (OpenRouter) fits 3.9× more context, which matters for long documents and agents.