GPT-5.4 Pro vs Qwen3.6 Max
Full profiles: GPT-5.4 Pro (OpenAI) and Qwen3.6 Max (Alibaba): every benchmark, price and rank.
Updated · Data from 6 sources
As of 27 Oct 2026, GPT-5.4 Pro and Qwen3.6 Max (max) are neck and neck on quality (63.2 vs 62.2 on the BenchLeader Index, inside the ±4.7 margin of error), and Qwen3.6 Max (max) costs 23× less, which makes it the better buy for most workloads. Pick GPT-5.4 Pro for reasoning (73 vs 58); pick Qwen3.6 Max for instruction following (75 vs 63).
GPT-5.4 Pro benchmarks 
Qwen3.6 Max benchmarks Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Reasoning | GPT-5.4 Pro | 73 vs 58 category score; SimpleBench 74.1% vs 63.0% | High |
| Knowledge | GPT-5.4 Pro | 74 vs 62 category score | High |
| Following instructions | Qwen3.6 Max | 75 vs 63 category score | High |
| Chat quality (human preference) | – | Only Qwen3.6 Max has an Arena rating (65 category score) | Not comparable |
| Lowest cost | Qwen3.6 Max | $2.92 vs $67.50 per 1M blended; agentic coding session $0.109 vs $2.52 | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- GPT-5.4 Pro and Qwen3.6 Max (max) are neck and neck on quality (63.2 vs 62.2, within the ±4.7 margin).
- GPT-5.4 Pro is stronger in knowledge, multimodal, reasoning.
- Qwen3.6 Max (max) is stronger in instruction following, agents & tools, coding, composite, human preference, long context, maths.
- Qwen3.6 Max (max) is 23× cheaper ($2.92 vs $67.50 per 1M blended).
At a glance
| Spec | GPT-5.4 Pro | Qwen3.6 Max |
|---|---|---|
| Developer | OpenAI | Alibaba |
| Released | 4 Mar 2026 | 20 Apr 2026 |
| Context window (tokens) | 1,050,000 | 262,144 |
| Compared at | GPT-5.4 Pro | Qwen3.6 Max (max) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $30.00 / $180.00 / – | $1.30 / $7.80 / $0.130 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | GPT-5.4 Pro | Qwen3.6 Max (max) | Leader |
|---|---|---|---|
| Reasoning | 72.5 | 58.1 | GPT-5.4 Pro |
| Coding | – | 56.6 | not comparable |
| Agents & tools | – | 72.1 | not comparable |
| Maths | – | 64.9 | not comparable |
| Knowledge | 73.6 | 62.1 | GPT-5.4 Pro |
| Instruction following | 63.2 | 74.6 | Qwen3.6 Max |
| Human preference | – | 64.8 | not comparable |
| Multimodal | 69.8 | – | not comparable |
| Long context | – | 65.2 | not comparable |
| Composite | – | 60.9 | not comparable |
Where sources disagree
Across the 2 benchmarks both have, GPT-5.4 Pro leads 2, Qwen3.6 Max leads 0.
| Overall index | GPT-5.4 Pro | Qwen3.6 Max (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 63.2 | 62.2 | neck and neck |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | GPT-5.4 Pro | Qwen3.6 Max (max) |
|---|---|---|
| Blended price per 1M tokens | $67.50 | $2.92 |
| Input / output per 1M | $30.00 / $180.00 | $1.30 / $7.80 |
| Cached input per 1M | – | $0.130 |
| Price per task: chat reply | $0.066 | $0.0029 |
| Price per task: summarize a 30-page report | $0.468 | $0.020 |
| Price per task: code edit | $0.450 | $0.019 |
| Price per task: agentic coding session | $2.52 | $0.109 |
| Price per task: structured extraction | $0.096 | $0.0042 |
Each model is compared at its best-ranked reasoning-effort setting. GPT-5.4 Pro scores 63.2; GPT-5.4 Pro (xhigh) scores 61.8. Qwen3.6 Max (max) scores 62.2. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
48 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | GPT-5.4 Pro | Qwen3.6 Max (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 63.2 | 62.2 |
| Instruction following score | BenchLeader | 63.2 | 74.6 |
| Knowledge score | BenchLeader | 73.6 | 62.1 |
| Multimodal score | BenchLeader | 69.8 | – |
| Reasoning score | BenchLeader | 72.5 | 58.1 |
| Agents & tools score | BenchLeader | – | 72.1 |
| Coding score | BenchLeader | – | 56.6 |
| Composite score | BenchLeader | – | 60.9 |
| Human preference score | BenchLeader | – | 64.8 |
| Long context score | BenchLeader | – | 65.2 |
| Maths score | BenchLeader | – | 64.9 |
| Blended price $/M | BenchLeader | $67.50 | $2.92 |
| Context window | BenchLeader | 1,050,000 tokens | 262,144 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 94.6%† | 87.4% |
| Humanity's Last Exam | Scale AI / CAIS | 44.3% | – |
| SimpleBench | SimpleBench | 74.1% | 63.0% |
| Arena Hard Prompts | Arena (LMArena) | – | 1483 |
| GPQA Diamond (AA) | Artificial Analysis | – | 88.8% |
| Humanity's Last Exam (AA) | Artificial Analysis | – | 30.8% |
| ARC-AGI-1 | ARC Prize | 94.5%† | – |
| ARC-AGI-2 | ARC Prize | 83.3%† | – |
| CritPt | Artificial Analysis | 30.0%† | 3.7% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 59.0%† | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | – | 19.0%† |
| LMCA | Epoch AI Benchmarking Hub | – | 42.5% |
| DTBench | Epoch AI Benchmarking Hub | – | 87.2% |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | – | 76.7% |
| WeirdML | WeirdML | 57.4%† | – |
| Arena Coding | Arena (LMArena) | – | 1512 |
| Arena WebDev | Arena (LMArena) | – | 1482 |
| SWE-bench (Vals) | Vals AI | – | 72.8% |
| Agents & tools | |||
| Terminal-Bench Hard | Artificial Analysis | – | 43.9% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 95.9% |
| Terminal-Bench 2.0 (Vals) | Vals AI | – | 51.7% |
| Vending-Bench 2 | Andon Labs | – | 4254.2 |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 82.5%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 58.5%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | – | 91.1% |
Show all 48 benchmarks
| Metric | Publisher | GPT-5.4 Pro | Qwen3.6 Max (max) |
|---|---|---|---|
| Maths | |||
| MathArena Apex | MathArena | 69.8%† | – |
| Arena Maths | Arena (LMArena) | – | 1476 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 46.3%† | 52.0% |
| AA-Omniscience | Artificial Analysis | – | 9.2 |
| CorpFin | Vals AI | – | 66.5% |
| MultiNRC | Scale AI SEAL | 62.3% | – |
| CaseLaw v2 | Vals AI | – | 47.9% |
| AA-Omniscience: accuracy | Artificial Analysis | – | 37.9% |
| AA-Omniscience: non-hallucination | Artificial Analysis | – | 53.8% |
| Instruction following | |||
| IFBench | Artificial Analysis | – | 76.6% |
| MultiChallenge | Scale AI SEAL | 69.2% | – |
| TutorBench | Scale AI SEAL | 56.6% | – |
| Arena Instruction Following | Arena (LMArena) | – | 1452 |
| Human preference | |||
| Arena Text | Arena (LMArena) | – | 1461 |
| Arena Creative Writing | Arena (LMArena) | – | 1438 |
| Arena Multi-turn | Arena (LMArena) | – | 1468 |
| Arena Longer Queries | Arena (LMArena) | – | 1474 |
| Multimodal | |||
| VISTA | Scale AI SEAL | 53.9% | – |
| Long context | |||
| AA-LCR | Artificial Analysis | – | 80.7% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 159.6 | 149.4 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | – | 28.4 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 14.8% | – |
| MASK | Scale AI SEAL | 91.7% | – |
Data as of 2026-10-27. Best configuration of each model; every score links to its source on the model pages.
What changed
- GPT-5.4 Pro: 1 score revision.
- Qwen3.6 Max (max): BenchLeader Index 62.8 → 62.3.
- GPT-5.4 Pro: BenchLeader Index 62.1 → 63.6.
- GPT-5.4 Pro: 1 new result (MultiNRC).
- GPT-5.4 Pro: BenchLeader Index 64.0 → 62.1.
- GPT-5.4 Pro: BenchLeader Index 63.4 → 64.0.
- GPT-5.4 Pro: 1 new result (TutorBench).
- GPT-5.4 Pro: BenchLeader Index 64.0 → 63.4.
- GPT-5.4 Pro: BenchLeader Index 62.6 → 64.0.
- GPT-5.4 Pro: 1 new result (MultiChallenge).
Related comparisons
- GPT-5.4 Pro vs Claude Opus 5.5
- GPT-5.4 Pro vs Claude Fable 5.1
- GPT-5.4 Pro vs GPT-6 Astra
- Qwen3.6 Max vs Claude Opus 5.5
- Qwen3.6 Max vs Claude Fable 5.1
- Qwen3.6 Max vs GPT-6 Astra
- Qwen3.6 Max vs Muse Wonder Full
- GPT-5.3-Codex vs Qwen3.6 Max
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
GPT-5.4 Pro vs Qwen3.6 Max: questions
- Is GPT-5.4 Pro better than Qwen3.6 Max?
- GPT-5.4 Pro and Qwen3.6 Max (max) are neck and neck on quality (63.2 vs 62.2, within the ±4.7 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-27, so check the category scores for your use.
- Which is cheaper, GPT-5.4 Pro or Qwen3.6 Max, and will that change?
- Qwen3.6 Max is cheaper today: $2.92 against $67.50 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Which is better for long documents?
- GPT-5.4 Pro accepts more context: 1,050,000 against 262,144 tokens for Qwen3.6 Max.
- Do independent benchmarks agree on GPT-5.4 Pro vs Qwen3.6 Max?
- Across the 2 benchmarks both have, GPT-5.4 Pro leads 2, Qwen3.6 Max leads 0.
- When was Qwen3.6 Max released?
- Qwen3.6 Max was released on 20 Apr 2026.