Gemma 4: Google's Open Source Model Family
Gemma 4 - Google's latest open source AI models built from Gemini 3 research. Download, API access, and performance benchmarks.
·2 min read
Share
Gemma 4 is Google's latest open source model family, built from Gemini 3 research and technology. It maximizes intelligence-per-parameter and is designed for efficiency across devices.
Models
| Model | Description | Best For |
|---|---|---|
| Gemma 4 27B | Flagship model | Complex reasoning, coding |
| Gemma 4 26B | Slightly smaller | Balanced performance |
| Gemma 4 E4B | Efficient 4B params | Resource-constrained devices |
| Gemma 4 E2B | Most compact | Mobile/IoT |
Key Features
- Agentic workflows — Native function calling support for autonomous agents
- Multimodal reasoning — Audio and visual understanding
- 140 languages — Multilingual with cultural context
- Fine-tuning — Custom training with your own data
Benchmark Performance
| Benchmark | Gemma 4 27B | Gemma 4 26B | Gemma 4 E4B | Gemma 3 27B |
|---|---|---|---|---|
| MMMLU | 85.2% | 82.6% | 69.4% | 67.6% |
| MMMU Pro | 76.9% | 73.8% | 52.6% | 49.7% |
| AIME 2026 Math | 89.2% | 88.3% | 42.5% | 20.8% |
| LiveCodeBench | 80.0% | 77.1% | 52.0% | 29.1% |
| GPQA | 84.3% | 82.3% | 58.6% | 42.4% |
| τ2-bench (Agent) | 86.4% | 85.5% | 57.5% | 6.6% |
Where to Get It
Download
- Hugging Face — Weights & fine-tuning
- Kaggle — Alternative download
API Access
- Google AI Studio — Free tier available
- Vertex AI — Enterprise deployment
- Third-party providers — RunPod, Modal, etc.
Local Running
- Ollama —
ollama run gemma4(check availability) - llama.cpp — For local inference
- LM Studio — Desktop UI
Use Cases
- Code generation & debugging
- Mathematical reasoning
- Agentic tool use
- Research & benchmarking
- Fine-tuning for domain-specific apps
Resources
Related
- LLM API Pricing — Compare all LLM costs
- Google Gemini API Pricing — Gemini pricing
- Best AI Tools by Category — More AI tools
Last updated: April 2026
Related Resources
ResourceNew AI Model Releases: How to Track, Test, and ChooseA practical guide to tracking new AI model releases, reading model announcements critically, testing upgrades, and deciding when to migrate.ResourceLLM API Pricing ComparisonCompare LLM API pricing from TLDL's verified source dataset. Pending providers are tracked but excluded until official pricing is verified.BlogKimi K3 Launches: What the 2.8T Open Model ChangesKimi K3 arrives with 2.8 trillion parameters, native vision, a 1M-token context window, long-horizon coding, and aggressive cache pricing. Here is what matters beyond the headline.ResourceKimi K3 API PricingKimi K3 API pricing from verified first-party sources, including cached and uncached input, output, 1M-token context, cost examples, and API limits.
Read more TLDL briefings
Browse website-published AI podcast briefings, resource updates, and research notes.
Published on TLDL. Follow by RSS if you want updates without another inbox.