Pricingby GlyphGemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output.
Ways to buy 19
| Way to buy | Input | Output | In / out per 1M | Cached | Vs cheapest |
|---|---|---|---|---|---|
fp4 · cached $0.1 | $0.1 | $0.34 | $0.1 / $0.34Cheapest | $0.1 | Cheapest |
fp4 · cached $0.09 | $0.12 | $0.36 | $0.12 / $0.36+12% vs cheapest | $0.09 | +12% vs cheapest |
fp4 · cached $0.012 | $0.12 | $0.37 | $0.12 / $0.37+14% vs cheapest | $0.012 | +14% vs cheapest |
fp8-flex, fp8 | $0.16 | $0.32 | $0.16 / $0.32+25% vs cheapest | n/a | +25% vs cheapest |
bf16 · cached $0.14 | $0.14 | $0.4 | $0.14 / $0.4+28% vs cheapest | $0.14 | +28% vs cheapest |
| DekaLLM Default region · cached $0.05 | $0.14 | $0.4 | $0.14 / $0.4+28% vs cheapest | $0.05 | +28% vs cheapest |
Default region | $0.14 | $0.4 | $0.14 / $0.4+28% vs cheapest | n/a | +28% vs cheapest |
bf16 | $0.14 | $0.4 | $0.14 / $0.4+28% vs cheapest | n/a | +28% vs cheapest |
| NextBit fp4 · cached $0.06 | $0.15 | $0.4 | $0.15 / $0.4+33% vs cheapest | $0.06 | +33% vs cheapest |
fp8 · cached $0.06 | $0.15 | $0.4 | $0.15 / $0.4+33% vs cheapest | $0.06 | +33% vs cheapest |
turbo, fp4 · cached $0.075 | $0.135 | $0.51 | $0.135 / $0.51+43% vs cheapest | $0.075 | +43% vs cheapest |
fp8 | $0.2 | $0.4 | $0.2 / $0.4+56% vs cheapest | n/a | +56% vs cheapest |
Default region · cached $0.1 | $0.34 | $0.1 | $0.34 / $0.1+75% vs cheapest | $0.1 | +75% vs cheapest |
fp8-priority, fp8 | $0.3 | $0.6 | $0.3 / $0.6+134% vs cheapest | n/a | +134% vs cheapest |
| PrimeIntellect fp4 · cached $0.2 | $0.4 | $0.67 | $0.4 / $0.67+192% vs cheapest | $0.2 | +192% vs cheapest |
| Io Net Default region · cached $0.19 | $0.38 | $1.15 | $0.38 / $1.15+258% vs cheapest | $0.19 | +258% vs cheapest |
Default region | $0.38 | $1.15 | $0.38 / $1.15+258% vs cheapest | n/a | +258% vs cheapest |
| ModelRun fp4 · cached $0.2 | $0.75 | $1 | $0.75 / $1+408% vs cheapest | $0.2 | +408% vs cheapest |
fp8 · cached $0.25 | $0.75 | $1 | $0.75 / $1+408% vs cheapest | $0.25 | +408% vs cheapest |
USD per 1M tokens. Regions are separate routes.
Through a gateway
Your own numbers100M input and 20M output tokens a month: $16.80 direct.
Kong AI Gateway$16.80+0%
LiteLLM$16.80+0%
Vercel AI Gateway$16.80+0%
Requesty$17.64+5%
Eden AI$17.72+5.5%
OpenRouter$17.72+5.5%
Cloudflare AI GatewayNot offeredNot offered
HeliconeNot offeredNot offered
Neon AI GatewayNot offeredNot offered
A dash: not on that gateway's own model list. No published token fee: Martian Gateway, Portkey, TrueFoundry AI Gateway.
Related models
| In / out per 1M | Cheapest route | |||||||
|---|---|---|---|---|---|---|---|---|
Google · 131K | $1.5 | $9 | $1.5 / $9 | $3.38 | 131K | Lab list price | · | Gemini Omni 1.2 Flash is Google's native multimodal video generation and editing model, the successor to Gemini Omni 1.1 Flash. |
Google · 131K | $1.5 | $9 | $1.5 / $9 | $3.38 | 131K | Lab list price | · | Gemini Omni 1.1 Flash is Google's native multimodal video generation and editing model, the generally available successor to Gemini Omni Flash Preview. |
Google · 1MCheapest: $2 on Google Gemini API (flex) | $2 | $10 | $2 / $10 | $4 | 1M | $2 on Google Gemini API (flex) | · | Gemini 4 Argon is Google's flagship reasoning model, built for complex software engineering, deep mathematical and scientific problem solving, and autonomous multi-step agents. |
Google · 66K | $1.5 | $7.5 | $1.5 / $7.5 | $3 | 66K | Lab list price | · | Nano Banana 2.1 (Gemini Nano Banana 2.1) is Google's image generation and editing model on the Flash tier, succeeding Nano Banana 2 and Nano Banana Pro. |
DeepSeek · 1MCheapest: $0.081 on Particle AI (us-flex) | $0.15 | $0.6 | $0.15 / $0.6 | $0.262 | 1M | $0.081 on Particle AI (us-flex) | · | DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash, served by DeepSeek under the deepseek-flash API model name. |
DeepSeek · 1M | $0.045 | $0.3 | $0.045 / $0.3 | $0.109 | 1M | on InferenceNet (fp8) | · | DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. |
Anthropic · 1M | $0.1 | $0.5 | $0.1 / $0.5 | $0.2 | 1M | Lab list price | · | Claude Haiku 5.5 is Anthropic's small, fast model for high-volume, cost-sensitive work such as summarization, subagents, and browser use. |
Cohere · 128K | $0.12 | Free | $0.12 / Free | $0.09 | 128K | Lab list price | · | Embed 5 Pro is the higher-accuracy tier of Cohere's Embed 5 family of multimodal embedding models (about 2.4B parameters). |
About Gemma 4 31B
Takes image, text, video in and returns text. First listed Sep 23, 2026. Up to 16K output tokens. Blended price $0.16 per 1M tokens at 3 input to 1 output. Model id gemma-4-31b-it.
Price history
No list-price history: this model is sold by hosts rather than its lab.
Questions
How much does Gemma 4 31B cost?
$0.1 per million input tokens and $0.34 per million output tokens, with cached input at $0.1.
What is the context window of Gemma 4 31B?
262K tokens, with up to 16K tokens of output.
Where is Gemma 4 31B cheapest?
Of 19 routes tracked, CoreWeave is cheapest at $0.1 input and $0.34 output per million tokens.
Has Gemma 4 31B's price changed?
Yes. The output price went from $0.182 to $0.204, first seen Nov 6, 2026.