Neither wins outright: Google: Gemma 4 31B is the pick when you need seed (reproducible sampling), Qwen: Qwen3.5-9B (batch) when cost matters at scale.
| Spec | Google: Gemma 4 31B | Qwen: Qwen3.5-9B (batch) |
|---|---|---|
| Provider | Qwen | |
| Input price (per 1M tokens) | $0.20 | $0.17 |
| Output price (per 1M tokens) | $0.40 | $0.25 |
| Blended price (3:1 input/output) | $0.25 | $0.19 |
| Cheaper than … of all models we track | 66% | 70% |
| Context window | 262K tokens (~393 pages) | 262K tokens (~393 pages) |
| Larger context than … of all models | 48% | 48% |
| Max output per response | 16K tokens (~12,300 words) | 236K tokens (~176,900 words) |
| Input types | image, text, video | text, image, video |
| Output types | text | text |
| Free variant | Yes | No |
| Listed since | 2026-04-02 | 2026-03-10 |
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) | Google: Gemma 4 31B | Qwen: Qwen3.5-9B (batch) | Difference |
|---|---|---|---|
| Customer-support chatbot 50,000 conversations — each a 6-turn conversation (1,500 in / 400 out tokens) | $23.00 | $17.75 | Qwen: Qwen3.5-9B (batch) saves $5.25 |
| RAG search over documents 20,000 questions — each a question answered from ~8 retrieved passages (6,000 in / 500 out tokens) | $28.00 | $22.90 | Qwen: Qwen3.5-9B (batch) saves $5.10 |
| 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) | $28.80 | $23.40 | Qwen: Qwen3.5-9B (batch) saves $5.40 |
| Batch summarisation 10,000 documents — each a 6,000-word report summarised to a page (8,000 in / 600 out tokens) | $18.40 | $15.10 | Qwen: Qwen3.5-9B (batch) saves $3.30 |
| Long-document analysis 500 documents — each a 150-page contract or codebase read in one pass (100,000 in / 2,000 out tokens) | $10.40 | $8.75 | Qwen: Qwen3.5-9B (batch) saves $1.65 |
| Capability | Google: Gemma 4 31B | Qwen: Qwen3.5-9B (batch) | 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 | ✓ 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 | — | 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 |
|---|---|---|---|
| 2026-10-15 | Google: Gemma 4 31B | $0.20 | $0.40 |
| 2026-08-27 | Google: Gemma 4 31B | $0.09 | $0.34 |
| 2026-08-21 | Google: Gemma 4 31B | $0.10 | $0.34 |
| 2026-08-19 | Google: Gemma 4 31B | $0.09 | $0.34 |
Neither wins outright: Google: Gemma 4 31B is the pick when you need seed (reproducible sampling), Qwen: Qwen3.5-9B (batch) when cost matters at scale. In short: Google: Gemma 4 31B when you need seed (reproducible sampling), you need log-probabilities, you want to prototype at zero cost; Qwen: Qwen3.5-9B (batch) when cost matters at scale, you generate long outputs in one go.
Qwen: Qwen3.5-9B (batch) is cheaper: $0.19 vs $0.25 per 1M tokens at a typical 3:1 input/output mix, 24% less. For 50,000 support conversations a month that is $17.75 against $23.00.
Google: Gemma 4 31B costs $0.20 per 1M input tokens and $0.40 per 1M output tokens; Qwen: Qwen3.5-9B (batch) costs $0.17 input and $0.25 output. Output tokens are billed separately, so long answers shift the gap.
Both take 262K tokens per request — roughly 393 pages of text.
Tool calling — Google: Gemma 4 31B: yes, Qwen: Qwen3.5-9B (batch): yes. JSON-schema structured output — Google: Gemma 4 31B: yes, Qwen: Qwen3.5-9B (batch): yes.
Yes, both accept image input.
Google: Gemma 4 31B has a free variant (google/gemma-4-31b-it:free). Free variants usually carry tighter rate limits.
Google: Gemma 4 31B can write up to 16K tokens (about 12,300 words) per response; Qwen: Qwen3.5-9B (batch) up to 236K (about 176,900 words).
Google: Gemma 4 31B was listed on 2026-04-02; Qwen: Qwen3.5-9B (batch) on 2026-03-10.
Google: Gemma 4 31B: input price raised 55% ($0.09 → $0.20/1M), output price raised 15% ($0.34 → $0.40/1M) on 2026-10-15 (4 price changes since 2026-08-06). Qwen: Qwen3.5-9B (batch): price unchanged since we started tracking it on 2026-08-29.
Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function...
Full Google: Gemma 4 31B specs →Qwen3.5-9B is a multimodal foundation model from the Qwen3.5 family, designed to deliver strong reasoning, coding, and visual understanding in an efficient 9B-parameter architecture. It uses a unified vision-language design...
Full Qwen: Qwen3.5-9B (batch) specs →