GPT-4.1 vs Claude Haiku 4.5 vs Qwen3 32B
Claude Haiku 4.5 comes out ahead, 58 to 53 and 51 on our weighted score, though Qwen3 32B is 39% cheaper per token.
OpenAI
GPT-4.1
53/100- ECI136.8
- Price$2.00 / $8.00
- Context1.05M
- Our pick
Anthropic
Claude Haiku 4.5
58/100- ECI142.4
- Price$1.00 / $5.00
- Context200K
Alibaba (Qwen)
Qwen3 32B
51/100- ECI138.5
- Price$0.70 / $2.80
- Context131K
Claude Haiku 4.5 is our pick
Claude Haiku 4.5 is the better all-round choice, scoring 58/100 against GPT-4.1 (53) and Qwen3 32B (51). It leads on capability and inputs & features. GPT-4.1 wins on context window. Qwen3 32B wins on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
- CapabilityClaude Haiku 4.5Capabilities Index (ECI): Claude Haiku 4.5 142.4 · Qwen3 32B 138.5 · GPT-4.1 136.8
- Lowest priceQwen3 32BQwen3 32B $1.23 · Claude Haiku 4.5 $2.00 · GPT-4.1 $3.50 per 1M tokens (3:1 blend)
- Longest contextGPT-4.1GPT-4.1 1,047,576 · Claude Haiku 4.5 200,000 · Qwen3 32B 131,072 tokens
- Widest inputsGPT-4.1 and Claude Haiku 4.5GPT-4.1: Text, Images, PDFs · Claude Haiku 4.5: Text, Images, PDFs · Qwen3 32B: Text
- Self-hostingQwen3 32BPublishes downloadable weights
| Measure | Weight | GPT-4.1 | Claude Haiku 4.5 | Qwen3 32B |
|---|---|---|---|---|
| CapabilityCapabilities Index (ECI) | 50% | 61 | 69 | 64 |
| Price | 25% | 24 | 36 | 46 |
| Inputs & features | 15% | 70 | 80 | 35 |
| Context window | 10% | 61 | 32 | 24 |
| Overall | 100% | 53/100 | 58/100 | 51/100 |
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | |||
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | 136.8 | 142.4 (best) | 138.5 |
| ECI rank | #111 of 148 | #90 of 148 (best) | #106 of 148 |
| GPQA DiamondGraduate-level science questions | 66.9% | 71.2% (best) | 65.7% |
| FrontierMath Tiers 1–3Research-level mathematics | 6.0% | — | — |
| OTIS Mock AIME 2024–2025Competition mathematics | 38.3% | 66.7% | 66.9% (best) |
| SWE-bench VerifiedFixing real GitHub issues | 48.5% | — | — |
| SimpleQA VerifiedShort factual questions | 31.1% (best) | 13.2% | — |
| Price per million tokens | |||
| Input | $2.00 | $1.00 | $0.70 (best) |
| Output | $8.00 | $5.00 | $2.80 (best) |
| Cached input | $0.50 | $0.10 (best) | — |
| Blended (3:1) | $3.50 | $2.00 | $1.23 (best) |
| Long-context rate | Same rate | Same rate | Same rate |
| Price source | Official OpenAI API | Official Anthropic API | Official Alibaba API |
| Limits | |||
| Context window | 1,047,576 tokens (best) | 200,000 tokens | 131,072 tokens |
| Max output | 32,768 tokens | 64,000 tokens (best) | 16,384 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | Yes | Yes | No |
| PDFs | Yes | Yes | No |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | No | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | Yes | Yes | No |
| Availability | |||
| Weights | Proprietary | Proprietary | Open |
| API model ID | gpt-4.1 | claude-haiku-4-5 | qwen3-32b |
| API providers | 25 | 34 (best) | 14 |
| Released | Apr 14, 2025 | Oct 15, 2025 | Apr 29, 2025 |
| Knowledge cutoff | Apr 2024 | Feb 28, 2025 | Apr 2025 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
GPT-4.1$36.00
Claude Haiku 4.5$20.00
Qwen3 32B$12.60
Which should you choose?
Which is better: GPT-4.1, Claude Haiku 4.5 or Qwen3 32B?
Claude Haiku 4.5 is the better all-round choice, scoring 58/100 against GPT-4.1 (53) and Qwen3 32B (51). It leads on capability and inputs & features. GPT-4.1 wins on context window. Qwen3 32B wins on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.
Which is cheaper, GPT-4.1, Claude Haiku 4.5 or Qwen3 32B?
Qwen3 32B is cheaper at $0.70 input / $2.80 output per million tokens (official Alibaba API price). Claude Haiku 4.5 costs $1.00 input / $5.00 output per million tokens (official Anthropic API price); GPT-4.1 costs $2.00 input / $8.00 output per million tokens (official OpenAI API price). At a typical mix of three input tokens to one output token, that is $1.23 per million tokens for Qwen3 32B versus $2.00 for Claude Haiku 4.5 (1.6× as much) and $3.50 for GPT-4.1 (2.9× as much).
Which scores higher on benchmarks?
Claude Haiku 4.5 scores higher on the Capabilities Index (ECI): Claude Haiku 4.5 142.4 (#90 of 148), Qwen3 32B 138.5 (#106 of 148) and GPT-4.1 136.8 (#111 of 148). The confidence ranges of the top two overlap (139.5–144.3 vs 135.1–140.4), so treat the gap as small. On individual benchmarks: GPQA Diamond — Claude Haiku 4.5 71.2%, GPT-4.1 66.9%, Qwen3 32B 65.7%; OTIS Mock AIME 2024–2025 — Qwen3 32B 66.9%, Claude Haiku 4.5 66.7%, GPT-4.1 38.3%.
Which is better for coding?
There are no published SWE-bench Verified results for Claude Haiku 4.5 and Qwen3 32B yet, so there is no like-for-like coding score. On overall capability, Claude Haiku 4.5 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.
Which has the bigger context window?
GPT-4.1 has the largest context window at 1,047,576 tokens, against 200,000 for Claude Haiku 4.5 and 131,072 for Qwen3 32B. Maximum output per response: GPT-4.1 up to 32,768, Claude Haiku 4.5 up to 64,000, Qwen3 32B up to 16,384 tokens.
Which can read images, PDFs, audio or video?
GPT-4.1 accepts text, images and PDFs; Claude Haiku 4.5 accepts text, images and PDFs; Qwen3 32B accepts text. GPT-4.1 handles the widest range of inputs.
Are any of these open source?
Qwen3 32B publishes its weights and can be self-hosted; GPT-4.1 and Claude Haiku 4.5 is proprietary.
Which is newer?
Claude Haiku 4.5 is the newest, released Oct 15, 2025. Qwen3 32B came out Apr 29, 2025; GPT-4.1 came out Apr 14, 2025. Knowledge cutoff: GPT-4.1 Apr 2024, Claude Haiku 4.5 Feb 28, 2025, Qwen3 32B Apr 2025.
How do you decide the winner?
Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Oct 4, 2026.