Seed-OSS-36B-Instruct vs Qwen3 235B-A22B Thinking 2507 vs Claude Haiku 4.5
Too close to call on our weighted score (Qwen3 235B-A22B Thinking 2507 67, Claude Haiku 4.5 66, Seed-OSS-36B-Instruct 60). The right pick depends on what you value most.
ByteDance SeedSeed-OSS-36B-Instruct
60/100- ECI—
- Price—
- Context524K
Alibaba (Qwen)Qwen3 235B-A22B Thinking 2507
67/100- ECI143.8
- Price$0.30 / $2.90
- Context262K
AnthropicClaude Haiku 4.5
66/100- ECI142.2
- Price$1.00 / $5.00
- Context200K
Too close to call
It is close. Our weighted score puts them within 1 points (Qwen3 235B-A22B Thinking 2507 67/100, Claude Haiku 4.5 66/100, Seed-OSS-36B-Instruct 60/100), so choose by what matters most for your work: Qwen3 235B-A22B Thinking 2507 for raw capability and Seed-OSS-36B-Instruct for long inputs. The score weighs capability 67%, inputs & features 20%, context window 13%. Capability uses the average of shared benchmarks (GPQA Diamond and OTIS Mock AIME 2024–2025), because Seed-OSS-36B-Instruct has no Capabilities Index score yet.
- CapabilityQwen3 235B-A22B Thinking 2507Shared benchmarks: Qwen3 235B-A22B Thinking 2507 83.4% · Seed-OSS-36B-Instruct 69.5% · Claude Haiku 4.5 68.9%
- Lowest priceQwen3 235B-A22B Thinking 2507Qwen3 235B-A22B Thinking 2507 $0.95 · Claude Haiku 4.5 $2.00 per 1M tokens (3:1 blend) · Seed-OSS-36B-Instruct unpriced
- Longest contextSeed-OSS-36B-InstructSeed-OSS-36B-Instruct 524,288 · Qwen3 235B-A22B Thinking 2507 262,144 · Claude Haiku 4.5 200,000 tokens
- Widest inputsClaude Haiku 4.5Seed-OSS-36B-Instruct: Text · Qwen3 235B-A22B Thinking 2507: Text · Claude Haiku 4.5: Text, Images, PDFs
- Self-hostingSeed-OSS-36B-Instruct and Qwen3 235B-A22B Thinking 2507Publishes downloadable weights (Apache 2.0)
| Measure | Weight | Seed-OSS-36B-Instruct | Qwen3 235B-A22B Thinking 2507 | Claude Haiku 4.5 |
|---|---|---|---|---|
| CapabilityShared benchmarks | 67% | 70 | 83 | 69 |
| Inputs & features | 20% | 35 | 35 | 80 |
| Context window | 13% | 49 | 37 | 32 |
| Overall | 100% | 60/100 | 67/100 | 66/100 |
Left out because at least one model lacks the data: price. The remaining weights were rescaled.
Every spec in one table
Highlighted cells lead their row. Dashes mean the data is not published.
| Specification | Seed-OSS-36B-InstructByteDance Seed | Qwen3 235B-A22B Thinking 2507Alibaba (Qwen) | Claude Haiku 4.5Anthropic |
|---|---|---|---|
| Capability | |||
| Capabilities Index (ECI) | — | 143.8 (best) | 142.2 |
| ECI rank | — | #93 of 165 (best) | #102 of 165 |
| GPQA DiamondGraduate-level science questions | 71.5% | 80.1% (best) | 71.2% |
| OTIS Mock AIME 2024–2025Competition mathematics | 67.5% | 86.7% (best) | 66.7% |
| SimpleQA VerifiedShort factual questions | — | 40.4% (best) | 13.2% |
| Price per million tokens | |||
| Input | — | $0.30 (best) | $1.00 |
| Output | — | $2.90 (best) | $5.00 |
| Cached input | — | — | $0.10 |
| Blended (3:1) | — | $0.95 (best) | $2.00 |
| Long-context rate | — | Same rate | Same rate |
| Price source | — | Median of 11 providers | Official Anthropic API |
| Limits | |||
| Context window | 524,288 tokens (best) | 262,144 tokens | 200,000 tokens |
| Max output | 524,288 tokens (best) | 81,920 tokens | 64,000 tokens |
| Inputs and features | |||
| Text | Yes | Yes | Yes |
| Images | No | No | Yes |
| PDFs | No | No | Yes |
| Audio | No | No | No |
| Video | No | No | No |
| Reasoning | Yes | Yes | Yes |
| Tool calling | Yes | Yes | Yes |
| Structured output | No | No | Yes |
| Availability | |||
| Weights | OpenApache 2.0 | OpenApache 2.0 | Proprietary |
| API model ID | — | — | claude-haiku-4-5 |
| API providers | — | 12 | 48 (best) |
| Released | Aug 20, 2025 | Jul 25, 2025 | Oct 15, 2025 |
| Knowledge cutoff | Jul 2024 | — | Feb 28, 2025 |
What would a month cost?
Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.
- Seed-OSS-36B-Instruct—
- Qwen3 235B-A22B Thinking 2507$8.80
- Claude Haiku 4.5$20.00
Which should you choose?
Which is better: Seed-OSS-36B-Instruct, Qwen3 235B-A22B Thinking 2507 or Claude Haiku 4.5?
It is close. Our weighted score puts them within 1 points (Qwen3 235B-A22B Thinking 2507 67/100, Claude Haiku 4.5 66/100, Seed-OSS-36B-Instruct 60/100), so choose by what matters most for your work: Qwen3 235B-A22B Thinking 2507 for raw capability and Seed-OSS-36B-Instruct for long inputs. The score weighs capability 67%, inputs & features 20%, context window 13%. Capability uses the average of shared benchmarks (GPQA Diamond and OTIS Mock AIME 2024–2025), because Seed-OSS-36B-Instruct has no Capabilities Index score yet.
Which is cheaper, Seed-OSS-36B-Instruct, Qwen3 235B-A22B Thinking 2507 or Claude Haiku 4.5?
Qwen3 235B-A22B Thinking 2507 is cheaper at $0.30 input / $2.90 output per million tokens (median across 11 API providers). Claude Haiku 4.5 costs $1.00 input / $5.00 output per million tokens (official Anthropic API price). At a typical mix of three input tokens to one output token, that is $0.95 per million tokens for Qwen3 235B-A22B Thinking 2507 versus $2.00 for Claude Haiku 4.5 (2.1× as much). Seed-OSS-36B-Instruct has no published per-token price.
Which scores higher on benchmarks?
Not every model here has a Capabilities Index score, so we compare the average of shared benchmarks (GPQA Diamond and OTIS Mock AIME 2024–2025): Qwen3 235B-A22B Thinking 2507 83.4%, Seed-OSS-36B-Instruct 69.5% and Claude Haiku 4.5 68.9%. On individual benchmarks: GPQA Diamond — Qwen3 235B-A22B Thinking 2507 80.1%, Seed-OSS-36B-Instruct 71.5%, Claude Haiku 4.5 71.2%; OTIS Mock AIME 2024–2025 — Qwen3 235B-A22B Thinking 2507 86.7%, Seed-OSS-36B-Instruct 67.5%, Claude Haiku 4.5 66.7%.
Which is better for coding?
There are no published SWE-bench Verified results for Seed-OSS-36B-Instruct, Qwen3 235B-A22B Thinking 2507 and Claude Haiku 4.5 yet, so there is no like-for-like coding score. On overall capability, Qwen3 235B-A22B Thinking 2507 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?
Seed-OSS-36B-Instruct has the largest context window at 524,288 tokens, against 262,144 for Qwen3 235B-A22B Thinking 2507 and 200,000 for Claude Haiku 4.5. Maximum output per response: Seed-OSS-36B-Instruct up to 524,288, Qwen3 235B-A22B Thinking 2507 up to 81,920, Claude Haiku 4.5 up to 64,000 tokens.
Which can read images, PDFs, audio or video?
Seed-OSS-36B-Instruct accepts text; Qwen3 235B-A22B Thinking 2507 accepts text; Claude Haiku 4.5 accepts text, images and PDFs. Claude Haiku 4.5 handles the widest range of inputs.
Are any of these open source?
Seed-OSS-36B-Instruct and Qwen3 235B-A22B Thinking 2507 publishes its weights (Apache 2.0) and can be self-hosted; Claude Haiku 4.5 is proprietary.
Which is newer?
Claude Haiku 4.5 is the newest, released Oct 15, 2025. Seed-OSS-36B-Instruct came out Aug 20, 2025; Qwen3 235B-A22B Thinking 2507 came out Jul 25, 2025. Knowledge cutoff: Seed-OSS-36B-Instruct Jul 2024, Claude Haiku 4.5 Feb 28, 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 Nov 2, 2026.
Related comparisons
- Seed-OSS-36B-Instruct vs GPT-5 NanoOpenAI · ECI 139.3
- Seed-OSS-36B-Instruct vs DeepSeek-R1DeepSeek · ECI 138.8
- Seed-OSS-36B-Instruct vs Gemma 4 26B A4B ITGoogle · ECI 142.0
- Qwen3 235B-A22B Thinking 2507 vs GLM-4.7Z.ai (Zhipu) · ECI 143.7
- Qwen3 235B-A22B Thinking 2507 vs GPT-5 MiniOpenAI · ECI 145.4
- Qwen3 235B-A22B Thinking 2507 vs DeepSeek R1 0528DeepSeek · ECI 141.2
