Home › Compare › Claude Haiku 4.5 vs Qwen 3 8B
Claude Haiku 4.5 vs Qwen 3 8B
A straight cost comparison of Claude Haiku 4.5 (Anthropic) and Qwen 3 8B (Alibaba (Qwen)), per 1,000,000 tokens, verified November 2026.
Rate-by-rate comparison
| Metric | Claude Haiku 4.5 | Qwen 3 8B | Winner |
|---|---|---|---|
| Input / 1M | $1 | $0.05 | Qwen 3 8B |
| Cached input / 1M | $0.10 | $0.005 | Qwen 3 8B |
| Cache write / 1M | $1.25 |  | Qwen 3 8B |
| Output / 1M | $5 | $0.20 | Qwen 3 8B |
| Context window | 200,000 | 131,072 | Claude Haiku 4.5 |
| Batch API | Yes | Yes | Tie |
| Long-context tier | None | None |  |
For per-token metrics, the lower number wins. For context window and batch support, the higher or broader option wins. A dash means the provider does not publish that rate: a missing cache-write fee means writes are not billed per token, while a missing cached-input rate means cache hits bill at the full input rate.
Blended cost: 1M input + 250K output
| Model | Blended cost | Difference |
|---|---|---|
| Claude Haiku 4.5 | $2.25 | +$2.15 |
| Qwen 3 8B | $0.10 | Cheaper |
Ten workloads, two bills
The same two models, priced across ten workloads that differ only in their input:output ratio. Nothing else changes  no retries, no system prompt, no middleware.
| Workload | Claude Haiku 4.5 | Qwen 3 8B | Winner |
|---|---|---|---|
| Classification 100K reqs · 500 in / 10 out | $55 | $2.70 | Qwen 3 8B |
| Data extraction 10K reqs · 2K in / 200 out | $30 | $1.40 | Qwen 3 8B |
| Support chat 1K reqs · 4K in / 600 out | $7 | $0.32 | Qwen 3 8B |
| Code review 2K reqs · 8K in / 1.5K out | $31 | $1.40 | Qwen 3 8B |
| Summarisation 1K reqs · 20K in / 1K out | $25 | $1.20 | Qwen 3 8B |
| Agent loop 5K reqs · 12K in / 2K out | $110 | $5 | Qwen 3 8B |
| Translation 500 reqs · 15K in / 15K out | $45 | $1.88 | Qwen 3 8B |
| Long-document QA 200 reqs · 120K in / 800 out | $24.80 | $1.23 | Qwen 3 8B |
| Cached RAG 1K reqs · 50K in (45K cached) / 1K out | $14.55 | $0.677 | Qwen 3 8B |
| Nightly batch 10K reqs · 10K in / 2K out | $100 | $4.50 | Qwen 3 8B |
Claude Haiku 4.5 wins 0 of 10, Qwen 3 8B wins 10 of 10. The winner is set by the token mix, not by the headline rate.
Where the two bills cross over
Qwen 3 8B is cheaper on both sides of the mix: $0.05 vs $1 per million input tokens, and $0.20 vs $5 per million output. No input:output ratio reverses that, so on price alone the other model never becomes the better choice  it has to win on context, latency or output quality.
Cost at four scales
The same support-chat workload (4K input, 600 output per request) at four monthly volumes.
| Requests / month | Claude Haiku 4.5 | Qwen 3 8B | Difference |
|---|---|---|---|
| 100 | $0.70 | $0.032 | Qwen 3 8B cheaper by $0.668 |
| 1,000 | $7 | $0.32 | Qwen 3 8B cheaper by $6.68 |
| 10,000 | $70 | $3.20 | Qwen 3 8B cheaper by $66.80 |
| 100,000 | $700 | $32 | Qwen 3 8B cheaper by $668 |
What caching does to each bill
Caching only helps the part of the prompt that repeats. This is the support-chat workload (1,000 requests, 4,000 input tokens each) with an increasing share of that prompt served from a warm cache, written once on the first request.
| Cached share of the prompt | Claude Haiku 4.5 | Qwen 3 8B | Winner |
|---|---|---|---|
| 0% | $7 | $0.32 | Qwen 3 8B |
| 50% | $5.20 | $0.23 | Qwen 3 8B |
| 80% | $4.12 | $0.176 | Qwen 3 8B |
| 95% | $3.58 | $0.149 | Qwen 3 8B |
Both models publish a cached-read rate, so the discount is comparable: Claude Haiku 4.5 reads cache at $0.10 per million, Qwen 3 8B at $0.005.
Which one should you pick?
Claude Haiku 4.5 wins on none of the ten workloads above, while Qwen 3 8B wins on classification, data extraction, support chat, code review, summarisation, agent loop, translation, long-document qa, cached rag, nightly batch.
Beyond price, the two differ on input rate ($1 vs $0.05); output rate ($5 vs $0.20); cached-input rate ($0.10 vs $0.005); cache-write rate ($1.25 vs â€â€); context window (200,000 vs 131,072).
Batch mode is available on Claude Haiku 4.5 and Qwen 3 8B, which halves the nightly job. Both publish cached-input discounts, so a stable system prompt is worth caching on either.
FAQ
Is Claude Haiku 4.5 cheaper than Qwen 3 8B?
On a blended 1M-input-plus-250K-output workload, Qwen 3 8B is cheaper by 96% ($0.10 vs $2.25). The gap changes for cache-heavy prompts and for batch jobs, as the workload table shows.
Which handles longer prompts, Claude Haiku 4.5 or Qwen 3 8B?
Claude Haiku 4.5 has the larger context window (200,000 vs 131,072 tokens). Neither model publishes a long-context surcharge.
Do both models support prompt caching?
Claude Haiku 4.5 bills cached reads at $0.10 per 1M; Qwen 3 8B at $0.005.
Related comparisons
- Claude Sonnet 5 vs Claude Haiku 4.5
- Qwen 3 32B vs Qwen 3 8B
- GPT-6 Luna vs Claude Haiku 4.5
- All Anthropic pricing
- All Alibaba (Qwen) pricing