Compare / Qwen3.5-27B vs GPT-5.4 mini
Qwen3.5-27B vs GPT-5.4 mini
Which is cheaper, which is more capable, and which does better at each kind of work, from 16 results on 3 sources that measured both models.
Of the 16 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.
Results as of 9 October 2026 (catalogue release 2026-10-09-dd620e778ac5). See also Alibaba vs OpenAI, every model from both labs.
Qwen3.5-27B or GPT-5.4 mini? The short answer
They score about the same, so pick Qwen3.5-27B: it costs 2.3 times less per token.
- Qwen3.5-27B is cheaper: $0.74 against $1.69 per million tokens, 2.3 times less.
- GPT-5.4 mini scores higher on the Artificial Analysis Intelligence Index: 24.1 against 22.9.
- GPT-5.4 mini is newer: released 17 Mar 2026, against 24 Feb 2026 for Qwen3.5-27B.
- On third-party benchmarks, on the results both models have: GPT-5.4 mini does better for reasoning and knowledge, writing and sticking to the facts.
Head to head
Every fact side by side
| Qwen3.5-27B | GPT-5.4 mini | |
|---|---|---|
| Price per million tokens | $0.27 in · $2.16 out | $0.75 in · $4.50 out |
| Blended (3 in : 1 out) | $0.74 | $1.69 |
| Cheapest host, blended | $0.54 (Alibaba) | $1.69 (Azure) |
| Cached input, per million | – | $0.075 cached input |
| Intelligence Index | 22.9 | 24.1 |
| Spring Prompt overall | – | – |
| Output speed, tokens a second | – | – |
| Context window | 262,144 tokens | 400,000 tokens |
| Released | 24 Feb 2026 | 17 Mar 2026 (newer) |
| Weights | Open | Closed |
| Developer based in | China | the US |
Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.
Which is better for what
Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.
Every shared result
Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.
| Measure | Qwen3.5-27B | GPT-5.4 mini |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results | ||
| Overall · TextArena rating, higher is better 129 · 72 of 387 measured | 1,409 | 1,447high reasoning |
| Overall · Text factualityArena rating, higher is better 112 · 64 of 163 measured | 1,420 | 1,447high reasoning |
| Business, management and finance · TextArena rating, higher is better 128 · 53 of 380 measured | 1,408 | 1,460high reasoning |
| Artificial Analysis reported by Artificial Analysis · 3 results | ||
| Artificial Analysis Intelligence Indexindex score, higher is better 86 · 81 of 243 measured | 22.9 | 24.1xhigh reasoning |
| GPQA Diamond% of questions, higher is better 74 · 61 of 231 measured | 85.8% | 87.5%xhigh reasoning |
| Humanity's Last Exam% of questions, higher is better 95 · 82 of 242 measured | 23.9% | 28.1%xhigh reasoning |
| Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results | ||
| Hallucination rate · Vectara% of summaries, lower is better 80 · 18 of 104 measured | 12.1% | 5.5% |
| Answer rate · Vectara% of documents, higher is better 31 · 1 of 104 measured | 99.8% | 100.0% |
Show all 16 results (8 more)
| Measure | Qwen3.5-27B | GPT-5.4 mini |
|---|---|---|
| Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results | ||
| Creative writing · TextArena rating, higher is better 145 · 92 of 385 measured | 1,358 | 1,401high reasoning |
| Expert prompts · TextArena rating, higher is better 112 · 63 of 338 measured | 1,441 | 1,480high reasoning |
| Instruction following · TextArena rating, higher is better 122 · 77 of 387 measured | 1,401 | 1,434high reasoning |
| Writing, literature and language · TextArena rating, higher is better 135 · 75 of 386 measured | 1,380 | 1,423high reasoning |
| Artificial Analysis reported by Artificial Analysis · 4 results | ||
| IFBench% of instructions, higher is better 24 · 34 of 195 measured | 75.6% | 73.3%xhigh reasoning |
| Long-context reasoning (AA-LCR)% of questions, higher is better 63 · 70 of 233 measured | 77.7% | 77.0%xhigh reasoning |
| Terminal-Bench Hard% of tasks, higher is better 61 · 11 of 190 measured | 32.6% | 52.3%xhigh reasoning |
| τ²-bench telecom% of tasks, higher is better 28 · 65 of 193 measured | 93.9% | 83.3%xhigh reasoning |
Compare them with other models
Quick answers
Is Qwen3.5-27B better than GPT-5.4 mini?
They score about the same, so pick Qwen3.5-27B: it costs 2.3 times less per token.
Which is cheaper, Qwen3.5-27B or GPT-5.4 mini?
Qwen3.5-27B is cheaper: $0.74 against $1.69 per million tokens, 2.3 times less.
Which is better for agents and tool use, Qwen3.5-27B or GPT-5.4 mini?
GPT-5.4 mini ranks 56 of 164 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 145 for Qwen3.5-27B.
Which is better for reasoning and knowledge, Qwen3.5-27B or GPT-5.4 mini?
GPT-5.4 mini ranks 71 of 198 models on sale for hard questions across science, maths and general knowledge, against 82 for Qwen3.5-27B.
Which is better for writing, Qwen3.5-27B or GPT-5.4 mini?
GPT-5.4 mini ranks 58 of 188 models on sale for what people prefer in blind comparisons, and judged writing quality, against 98 for Qwen3.5-27B.
Which is better for sticking to the facts, Qwen3.5-27B or GPT-5.4 mini?
GPT-5.4 mini ranks 66 of 152 models on sale for summarising without inventing things, and factual answers, against 121 for Qwen3.5-27B.
Which has the bigger context window, Qwen3.5-27B or GPT-5.4 mini?
GPT-5.4 mini: 400,000 tokens, against 262,144 for Qwen3.5-27B.
Run Qwen3.5-27B and GPT-5.4 mini on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.