# Gemma

> Source: https://agentsindex.ai/gemma

Open models for generation, retrieval, and safety checks.

- Best for: Developers who need open models they can adapt and deploy across devices and cloud.

Starting price

Free

From [ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing), checked 28 October 2026

- MCP server Not stated
- API Yes
- Open source Yes Checked by us github.com/google-deepmind/gemma/blob/main/LICENSE, 15 November 2026 [Checked by us](https://github.com/google-deepmind/gemma/blob/main/LICENSE)
- Self-host Yes

[Docs](https://ai.google.dev/gemma/docs)

What we checked

- Pages last read: 15 November 2026
- Facts checked: 2 facts · From 2 sources
- AI answers: Too few read · 58 buyer-question answers read, November 2026

[How we check](https://agentsindex.ai/methodology) [Editorial policy](https://agentsindex.ai/editorial-policy#not-for-sale)

[Visit Gemma](https://agentsindex.ai/go/gemma) [Compare](https://agentsindex.ai/compare?cat=ai-model-providers&products=gemma)

## At a glance

- Category: [AI model providers](https://agentsindex.ai/categories/ai-model-providers)
- Tags: Local AI, Open source, Self-hosted

### For developers

- Works with: Android, iOS, Chrome, JAX, Keras, PyTorch and 11 more

## Gemma in our research

[All research](https://agentsindex.ai/blog)

Study 19 October 2026 – 16 November 2026

### [AI Tool Pricing Index, November 2026](https://agentsindex.ai/data/ai-tool-pricing-index)

Pricing

Free only

29 of the 361 tools we checked are free with no paid plan.

Edition price, checked 28 October 2026.

Pricing pages of the AI tools, November 2026 edition as of 16 November 2026 n = 361 tools checked

## What is Gemma?

Gemma is a family of open models for developers who need adaptable AI for generation, retrieval, and safety checks. It includes Gemma 4 for text, audio, and image input, EmbeddingGemma, and ShieldGemma 2, with support for on-device and cloud deployment. The stack works with JAX, Keras, PyTorch, TensorFlow, and vLLM, and is available through Kaggle Models, Hugging Face, and Google Cloud.

## What changed with Gemma 4: a real license, not a marketing one

Every Gemma release before this one shipped under the custom Gemma Terms of Use, which grants broad commercial rights but incorporates a Prohibited Use Policy by reference, includes a clause letting Google remotely restrict access it believes violates the agreement, and requires redistributors to pass the terms downstream ([ai.google.dev/gemma/terms](https://ai.google.dev/gemma/terms)). That page is still live and still governs Gemma 3 and the specialized variants. Gemma 4, released April 2, 2026, switched to plain Apache 2.0 — confirmed on its official model card ([ai.google.dev/gemma/docs/core/model_card_4](https://ai.google.dev/gemma/docs/core/model_card_4)), on Google's Open Source blog announcement ([opensource.googleblog.com/2026/03/gemma-4-expanding-the-gemmaverse-with-apache-20.html](https://opensource.googleblog.com/2026/03/gemma-4-expanding-the-gemmaverse-with-apache-20.html)), and independently in the `license` field of every Gemma 4 repo on Hugging Face's API (checked directly: google/gemma-4-31B-it, gemma-4-26B-A4B-it, gemma-4-12B-it, gemma-4-E4B-it, gemma-4-E2B-it all report `license:apache-2.0`). Apache 2.0 drops the acceptable-use flow-down and the unilateral kill-switch language — a meaningful difference if you were avoiding Gemma or Llama for exactly that reason. It does not grant trademark rights to the Gemma name.

## The model lineup and what it's actually good at

Gemma 4 ships in five sizes — E2B (2.3B effective params), E4B (4.5B), 12B (unified encoder-free architecture), 26B-A4B (25.2B total / 3.8B active, Mixture-of-Experts), and 31B dense — all multimodal (text, image; audio natively on E2B/E4B/12B), with 128K context on the small models and 256K on the 12B and up, and multilingual support across 140+ languages (model card: [ai.google.dev/gemma/docs/core/model_card_4](https://ai.google.dev/gemma/docs/core/model_card_4)). Independent tracking from Artificial Analysis puts the 31B reasoning variant at an Intelligence Index of 30, ranked #9 of 134 open-weight models it tracks in that size class, well above the median of 9 for similar-sized open models ([artificialanalysis.ai/models/gemma-4-31b](https://artificialanalysis.ai/models/gemma-4-31b)). The same source flags it as notably slow on raw output speed — 35.5 tokens/sec, ranked #66/134 — so it trades throughput for reasoning quality rather than winning on both. On the crowdsourced LMArena leaderboard, Google's own reporting places the 31B dense model at #3 among open models and the 26B MoE at #6 while activating under 4B parameters per token, based on blind human side-by-side ratings ([huggingface.co/blog/gemma4](https://huggingface.co/blog/gemma4)).

## Cost: free two ways, and one paid way that isn't cheap

There are three real paths to running Gemma, and they have different cost and privacy trade-offs. Self-hosting the open weights (via Hugging Face, Kaggle, Ollama, llama.cpp, or Unsloth) costs only your own compute — Gemma 4 weights are downloadable with no usage fee. Through Google's own Gemini API / AI Studio, Gemma 4 input, output and context-caching are all listed as free of charge with no paid tier offered at all — but the same pricing page states free-tier usage may be used to improve Google's products, which matters if you need a contractual no-training guarantee (checked directly on [ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing), page timestamped 13 August 2026). If you want a paid, presumably-no-training hosted endpoint instead, Vertex AI offers the Gemma 4 26B variant at $0.15 per million input tokens and $0.60 per million output tokens, with a $0.015 cache-hit rate (verified directly on [cloud.google.com/vertex-ai/generative-ai/pricing](https://cloud.google.com/vertex-ai/generative-ai/pricing) via its Gemma pricing table) — notably, only the 26B size appears with hosted pricing there, not the other four sizes.

## Ecosystem and maintenance signal

The reference google-deepmind/gemma GitHub repository (the inference library, not the weights) has 5,659 stars, 1,012 forks, and was last pushed to on 10 August 2026 — a week before this check — with 316 open issues, indicating active, ongoing maintenance rather than a one-off release dump (checked via GitHub API, [github.com/google-deepmind/gemma](https://github.com/google-deepmind/gemma)). Individual model weights show heavy real usage on Hugging Face: the gemma-4-31B-it and gemma-4-26B-A4B-it repos each report roughly 9.7–10 million downloads and thousands of likes (checked via Hugging Face's model API). Day-one tooling support is broad — Kaggle, Hugging Face, Keras, Ollama, PyTorch, JAX, Google AI Studio, Vertex AI, Android, LM Studio and Unsloth are all listed as supported access points on the official DeepMind models page ([deepmind.google/models/gemma](https://deepmind.google/models/gemma)).

## The defamation incident and what it tells you about intended use

In late October 2025, Google removed Gemma from AI Studio's web interface after Senator Marsha Blackburn said the model, asked a factual question about her, fabricated a detailed and false sexual-misconduct allegation with fake supporting article links (TechCrunch: [techcrunch.com/2025/11/02/google-pulls-gemma-from-ai-studio-after-senator-blackburn-accuses-model-of-defamation](https://techcrunch.com/2025/11/02/google-pulls-gemma-from-ai-studio-after-senator-blackburn-accuses-model-of-defamation/)). Google's public response was that it had seen non-developers using AI Studio's Gemma option to ask it factual, current-events-style questions, and said plainly: 'We never intended this to be a consumer tool or model, or to be used this way.' The removal was partial — Gemma stayed available to developers through the API and other deployment channels — and was framed by Google as a UI/positioning fix, not a model recall. It's a useful data point independent of whether Gemma 4 or later fixed the underlying hallucination behavior: this model family is positioned and marketed for developers building products, not for direct end-user question-answering, and a buyer building a public-facing Q&A or chat surface on Gemma should weight that accordingly.

## Safety tooling and known limitations

Google publishes a companion safety classifier, ShieldGemma 2, and offers content-safety filtering guidance in the model cards; the Gemma 4 model card documents its own ethics-and-safety evaluation approach and states standard prohibited-use categories (CSAM, dangerous instructions, explicit sexual content, hate speech) while explicitly telling integrators to add their own content-safety safeguards on top ([ai.google.dev/gemma/docs/core/model_card_4](https://ai.google.dev/gemma/docs/core/model_card_4)). Independent academic red-teaming has found the smaller Gemma 3 variants (1B) substantially more vulnerable to prompt-injection and jailbreak techniques than the larger ones in the same family, per a 2026 workshop paper evaluating multiple open models ([dl.acm.org/doi/10.1145/3803628.3807972](https://dl.acm.org/doi/10.1145/3803628.3807972)) — a pattern that is common across small open models generally, not unique to Gemma, but worth knowing if you're deploying the smallest on-device sizes without your own guardrail layer.

## Gemma price and terms

### Terms

- Free plan: Yes

The terms in full

On the Gemini Developer API the Gemma 4 table prices input, output and context caching as "Free of charge" on the Free Tier, with every Paid Tier cell reading "Not available"; tuning and Grounding with Google Search are "Not available" on both tiers, and Free Tier content is "Used to improve our products: Yes".

[![Gemma](https://agentsindex.ai/img/1024/screenshots/gemma/pricing.webp)](https://agentsindex.ai/img/1920/screenshots/gemma/pricing.webp)

Pricing page, captured 19 October 2026

[ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing), checked 28 October 2026

### All plans and limits

| Plan | Price | Main limits |
| --- | --- | --- |
| Gemma | Free | None stated |

Read 28 October 2026 from [ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing).

## What we checked on Gemma

Yes Partly No

What it costs to start Yes

### What the price counts

The vendor's own pricing table meters by tokens but charges nothing for Gemma 4: input, output and context caching all read "Free of charge" on the Free Tier while every Paid Tier cell reads "Not available", so there is no per-seat, per-use or quoted rate to pay for the model itself.

On the page: Gemma 4 section, pricing table header plus Input price row

Gemma 4 section, pricing table header plus Input price row [ai.google.dev/gemini-api/docs/pricing](https://ai.google.dev/gemini-api/docs/pricing) Checked 28 October 2026

How you build with it Yes

### Open source

Apache-2.0 licence in its public repository

> Apache License Version 2.0

LICENSE in github.com/google-deepmind/gemma [github.com/google-deepmind/gemma/blob/m…](https://github.com/google-deepmind/gemma/blob/main/LICENSE) Checked by us, 15 November 2026

## Gemma alternatives

- [Fireworks AI](https://agentsindex.ai/fireworks-ai) · Inference, model routing and fine-tuning for open models on one OpenAI-compatible endpoint, serverless or in your cloud. · From $0.05 per 1m input tokens Checked 6 Nov · HIPAA SOC 2 · [Visit](https://agentsindex.ai/go/fireworks-ai)
  - MCP server: partly
  - API: yes
  - Open source: no
  - Self-host: partly
- Gemma and Fireworks AI
  | Gemma | Free | 28 Oct |
  | --- | --- | --- |
  | [Fireworks AI](https://agentsindex.ai/fireworks-ai) | From $0.05 per 1m input tokens | 6 Nov |

## Questions about Gemma

- How much does Gemma cost?: Gemma is free to use. From ai.google.dev/gemini-api/docs/pricing, checked 28 October 2026.
- Does Gemma have an API or an MCP server?: Gemma has an API.
- Is Gemma open source, and can you host it yourself?: Gemma is open source. We checked github.com/google-deepmind/gemma/blob/main/LICENSE ourselves on 15 November 2026. You can host Gemma yourself.
- What have we checked on Gemma?: 2 facts, under these headings: What it costs to start, how you build with it.
- How do we check Gemma?: We hold 2 facts on Gemma, quoted from 2 pages. Facts last checked 15 November 2026. Price last checked 28 October 2026. Last change found 15 November 2026.

[How we check every listing](https://agentsindex.ai/methodology)

## Sources and changes

Last full check 28 October 2026

Sources 1

- Read 28 October 2026 · [ai.google.dev](https://ai.google.dev/gemini-api/docs/pricing) /gemini-api/docs/pricing

What changed 1

- 15 November 2026 What it does

## Gemma price history

We have read the price once so far, on 28 October 2026: free.

| Price | Plan | First seen | Last seen | Readings | Source |
| --- | --- | --- | --- | --- | --- |
| Free |   | 28 October 2026 | 28 October 2026 | 1 | [Pricing page, read 28 October 2026](https://ai.google.dev/gemini-api/docs/pricing) |

From our research on 28 October 2026. A date is when we read the page, not when the price changed.

## Similar to Gemma

- [ERIUS PHONE](https://agentsindex.ai/eriusphone) Rented cloud Android phones, each in its own container, that keep an agent's apps and sign-ins between sessions. From about $44 a month
- [TinyFish](https://agentsindex.ai/tinyfish) Web search, page fetching, cloud browsers and goal-driven web automation for AI agents, with search and fetch free. Free plan, then $0.002 per minute
- [Smithery](https://agentsindex.ai/smithery) Public MCP server registry, now run by Arcade, listing direct vendor MCP URLs after its proxy shut down. Free plan
- [UI Discovery](https://agentsindex.ai/uidiscovery) A library of real websites, iOS apps and dashboards with each one's palette, fonts and original build prompt. From $20 a month

## Is Gemma yours?

Free, and the page then carries the verified owner mark. You confirm it with an email at ai.google.dev. [Claim this listing](https://agentsindex.ai/claim/gemma?from=owner-line)

[Report an error](mailto:hello@agentsindex.ai?subject=Report%20an%20error%3A%20Gemma%20(gemma)&body=What%20is%20wrong%20on%20the%20page%2C%20and%20where%20on%20the%20product) if a fact is wrong.
