Neither wins outright: OpenAI: gpt-oss-20b (batch) is the pick when cost matters at scale, Qwen: Qwen3 VL 235B A22B Thinking when you need image input.
| Spec | OpenAI: gpt-oss-20b (batch) | Qwen: Qwen3 VL 235B A22B Thinking |
|---|---|---|
| Provider | OpenAI | Qwen |
| Input price (per 1M tokens) | $0.02 | $0.40 |
| Output price (per 1M tokens) | $0.11 | $4.00 |
| Blended price (3:1 input/output) | $0.05 | $1.30 |
| Cheaper than … of all models we track | 87% | 34% |
| Context window | 131K tokens (~197 pages) | 131K tokens (~197 pages) |
| Larger context than … of all models | 23% | 23% |
| Max output per response | 118K tokens (~88,500 words) | 33K tokens (~24,600 words) |
| Input types | text | text, image |
| Output types | text | text |
| Free variant | No | No |
| Listed since | 2025-08-05 | 2025-09-23 |
Monthly cost at list price for five common workloads. Token counts per unit are stated so you can scale them to your own traffic. A workload that does not fit a model's context window is marked rather than priced.
| Workload (per month) | OpenAI: gpt-oss-20b (batch) | Qwen: Qwen3 VL 235B A22B Thinking | Difference |
|---|---|---|---|
| Customer-support chatbot 50,000 conversations — each a 6-turn conversation (1,500 in / 400 out tokens) | $4.04 | $110 | OpenAI: gpt-oss-20b (batch) saves $106 |
| RAG search over documents 20,000 questions — each a question answered from ~8 retrieved passages (6,000 in / 500 out tokens) | $4.00 | $88.00 | OpenAI: gpt-oss-20b (batch) saves $84.00 |
| Coding / tool-using agent 2,000 tasks — each a multi-step task with tool calls and file context (60,000 in / 6,000 out tokens) | $4.22 | $96.00 | OpenAI: gpt-oss-20b (batch) saves $91.78 |
| Batch summarisation 10,000 documents — each a 6,000-word report summarised to a page (8,000 in / 600 out tokens) | $2.59 | $56.00 | OpenAI: gpt-oss-20b (batch) saves $53.41 |
| Long-document analysis 500 documents — each a 150-page contract or codebase read in one pass (100,000 in / 2,000 out tokens) | $1.31 | $24.00 | OpenAI: gpt-oss-20b (batch) saves $22.69 |
| Capability | OpenAI: gpt-oss-20b (batch) | Qwen: Qwen3 VL 235B A22B Thinking | What it lets you do |
|---|---|---|---|
| Tool / function calling | ✓ Yes | ✓ Yes | call your APIs and run agent loops |
| Structured output (JSON schema) | ✓ Yes | ✓ Yes | return JSON that validates against your schema |
| Reasoning / thinking mode | ✓ Yes | ✓ Yes | spend extra tokens thinking before answering hard problems |
| Image input | — | ✓ Yes | read screenshots, charts and scanned pages |
| Audio input | — | — | take speech or audio directly |
| File / PDF input | — | — | accept documents without your own parsing step |
| Image output | — | — | generate images, not just text |
| Built-in web search | — | — | answer from live web results |
| Seed (reproducible sampling) | ✓ Yes | ✓ Yes | repeat a generation for tests and evals |
| Log-probabilities | — | ✓ Yes | read token confidence for classification and scoring |
From each model's published parameters and input/output types. “—” means not advertised, not proven absent.
| Date | Model | Input / 1M | Output / 1M |
|---|---|---|---|
| 2026-09-23 | OpenAI: gpt-oss-20b (batch) | $0.02 | $0.11 |
| 2026-08-07 | Qwen: Qwen3 VL 235B A22B Thinking | $0.40 | $4.00 |
| 2026-08-06 | Qwen: Qwen3 VL 235B A22B Thinking | $0.98 | $3.95 |
| 2026-03-31 | Qwen: Qwen3 VL 235B A22B Thinking | $0.26 | $2.60 |
Neither wins outright: OpenAI: gpt-oss-20b (batch) is the pick when cost matters at scale, Qwen: Qwen3 VL 235B A22B Thinking when you need image input. In short: OpenAI: gpt-oss-20b (batch) when cost matters at scale, you generate long outputs in one go; Qwen: Qwen3 VL 235B A22B Thinking when you need image input, you need log-probabilities.
OpenAI: gpt-oss-20b (batch) is cheaper: $0.05 vs $1.30 per 1M tokens at a typical 3:1 input/output mix, 96% less. For 50,000 support conversations a month that is $4.04 against $110.
OpenAI: gpt-oss-20b (batch) costs $0.02 per 1M input tokens and $0.11 per 1M output tokens; Qwen: Qwen3 VL 235B A22B Thinking costs $0.40 input and $4.00 output. Output tokens are billed separately, so long answers shift the gap.
Both take 131K tokens per request — roughly 197 pages of text.
Tool calling — OpenAI: gpt-oss-20b (batch): yes, Qwen: Qwen3 VL 235B A22B Thinking: yes. JSON-schema structured output — OpenAI: gpt-oss-20b (batch): yes, Qwen: Qwen3 VL 235B A22B Thinking: yes.
Only Qwen: Qwen3 VL 235B A22B Thinking accepts image input; OpenAI: gpt-oss-20b (batch) is text-only for input.
No free variant is listed for either; both are pay-per-token.
OpenAI: gpt-oss-20b (batch) can write up to 118K tokens (about 88,500 words) per response; Qwen: Qwen3 VL 235B A22B Thinking up to 33K (about 24,600 words).
OpenAI: gpt-oss-20b (batch) was listed on 2025-08-05; Qwen: Qwen3 VL 235B A22B Thinking on 2025-09-23.
OpenAI: gpt-oss-20b (batch): input price cut 52% ($0.05 → $0.02/1M), output price cut 44% ($0.20 → $0.11/1M) on 2026-09-23. Qwen: Qwen3 VL 235B A22B Thinking: input price cut 59% ($0.98 → $0.40/1M), output price raised 1% ($3.95 → $4.00/1M) on 2026-08-07 (3 price changes since 2026-03-10).
gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...
Full OpenAI: gpt-oss-20b (batch) specs →Qwen3-VL-235B-A22B Thinking is a multimodal model that unifies strong text generation with visual understanding across images and video. The Thinking model is optimized for multimodal reasoning in STEM and math....
Full Qwen: Qwen3 VL 235B A22B Thinking specs →