Skip to content
English

Switch theme

Discover▾

Not sure where to start? This week’s hottest software, models and stacks in three minutes.

AI Apps▾

AI software people use: from assistants like ChatGPT to tools for coding, video and workflows, picked by purpose.

For Agents▾

Software for AI: tools that let your agent (Claude Code, Codex, Cursor…) browse, read docs, run code and remember.

Models▾

Choose models, estimate cost, self-host: prices, rankings, benchmarks, free tiers and compute.

Solutions▾

A goal or an error message? Ready-made stacks, steps and fixes by scenario.

Learn▾

A systematic course from zero to the frontier, with an algorithm library, must-read papers and tutorials.

Frontier▾

Daily AI news, weekly highlights, the latest papers and how AI is changing every discipline.

// Open-model VRAM table · family

Gemma: local deployment & VRAM

48 open-weight models from Google (87 official repos incl. quantised builds). At 4-bit they range from 0.7 GB (functiongemma-270m-it) to 23.5 GB (gemma-4-31B-it-qat-q4_0-unquantized).

Filter by my hardware → VRAM calculator

official orgs: googlelatest 2026-09-14memory = estimate

01Text LLM (12)

ModelParamsContextBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
functiongemma-270m-it268M — 1.1 GB0.7 GB 8GB GPUMac 16GB Commercial with conditions2025-10-08
vaultgemma-1b1.04B — 2.8 GB1.2 GB 8GB GPUMac 16GB Commercial with conditions2025-09-05
gemma-3-270m-it268M — 1.1 GB0.7 GB 8GB GPUMac 16GB Commercial with conditions2025-07-30
t5gemma-b-b-ul2591M — 1.8 GB0.9 GB 8GB GPUMac 16GB Commercial with conditions2025-06-19
t5gemma-s-s-prefixlm313M — 1.2 GB0.7 GB 8GB GPUMac 16GB Commercial with conditions2025-06-19
t5gemma-s-s-ul2 ↗313M — 1.2 GB0.7 GB 8GB GPUMac 16GB Commercial with conditions2025-06-19
medgemma-27b-text-it27B — 59.9 GB18.5 GB RTX 4090 24GBMac 36GB Commercial with conditions2025-05-19
gemma-3-1b-it1B — 2.7 GB1.2 GB 8GB GPUMac 16GB Commercial with conditions2025-03-10
gemma-2-2b-jpn-it2.61B — 6.3 GB2.2 GB 8GB GPUMac 16GB Commercial with conditions2024-09-25
gemma-2-2b-it2.61B — 6.3 GB2.2 GB 8GB GPUMac 16GB Commercial with conditions2024-07-16
gemma-2-9b-it9.24B — 20.8 GB6.7 GB 8GB GPUMac 16GB Commercial with conditions2024-06-24
gemma-2-27b-it27.2B — 60.4 GB18.7 GB RTX 4090 24GBMac 36GB Commercial with conditions2024-06-24

02Vision-language (23)

ModelParamsContextBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
diffusiongemma-26B-A4B-it25.8B / 4B 256K 57.9 GB18.3 GB RTX 4090 24GBMac 36GB Commercial OK2026-06-09
gemma-4-26B-A4B-it-qat-q4_0-unquantized25.8B / 4B 256K 57.8 GB18.3 GB RTX 4090 24GBMac 36GB Commercial OK2026-04-29
gemma-4-31B-it-qat-q4_0-unquantized31.3B 256K 71.5 GB23.5 GB RTX 5090 32GBMac 64GB Commercial OK2026-04-28
gemma-4-26B-A4B-it25.8B / 4B 256K 57.8 GB18.3 GB RTX 4090 24GBMac 36GB Commercial OK2026-03-11
gemma-4-31B-it31.3B 256K 71.5 GB23.5 GB RTX 5090 32GBMac 64GB Commercial OK2026-03-11
medgemma-1.5-4b-it4.3B — 10.0 GB3.4 GB 8GB GPUMac 16GB Commercial with conditions2026-01-07
t5gemma-2-270m-270m786M — 2.2 GB1.0 GB 8GB GPUMac 16GB Commercial with conditions2025-10-25
t5gemma-2-1b-1b2.12B — 5.2 GB1.9 GB 8GB GPUMac 16GB Commercial with conditions2025-10-25
t5gemma-2-4b-4b7.51B — 17.0 GB5.5 GB 8GB GPUMac 16GB Commercial with conditions2025-10-25
medgemma-27b-it27.4B — 60.9 GB18.8 GB RTX 4090 24GBMac 36GB Commercial with conditions2025-07-09
gemma-3n-E2B-it5.44B — 12.5 GB4.1 GB 8GB GPUMac 16GB Commercial with conditions2025-06-12
gemma-3n-E4B-it7.85B — 17.8 GB5.7 GB 8GB GPUMac 16GB Commercial with conditions2025-06-03
medgemma-4b-it4.3B — 10.0 GB3.4 GB 8GB GPUMac 16GB Commercial with conditions2025-05-19
gemma-3-12b-it-qat-q4_0-unquantized12.2B — 27.3 GB8.6 GB 12GB GPUMac 16GB Commercial with conditions2025-04-08
gemma-3-12b-it12.2B — 27.3 GB8.6 GB 12GB GPUMac 16GB Commercial with conditions2025-03-01
gemma-3-27b-it27.4B — 60.9 GB18.8 GB RTX 4090 24GBMac 36GB Commercial with conditions2025-03-01
gemma-3-4b-it4.3B — 10.0 GB3.4 GB 8GB GPUMac 16GB Commercial with conditions2025-02-20
paligemma2-3b-pt-2243.03B — 7.2 GB2.5 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21
paligemma2-3b-pt-4483.03B — 7.2 GB2.5 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21
paligemma2-3b-ft-docci-4483.03B — 7.2 GB2.5 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21
paligemma2-10b-ft-docci-448 ↗9.66B — 21.8 GB6.9 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21
paligemma2-3b-mix-2243.03B — 7.2 GB2.5 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21
paligemma2-3b-mix-4483.03B — 7.2 GB2.5 GB 8GB GPUMac 16GB Commercial with conditions2024-11-21

03Omni / audio LLM (6)

ModelParamsContextBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
gemma-4-12B-it-qat-q4_0-unquantized12B 256K 27.7 GB9.3 GB 12GB GPUMac 16GB Commercial OK2026-06-04
gemma-4-12B-it12B 256K 27.7 GB9.3 GB 12GB GPUMac 16GB Commercial OK2026-05-23
gemma-4-E4B-it-qat-q4_0-unquantized7.94B 128K 18.1 GB5.9 GB 8GB GPUMac 16GB Commercial OK2026-04-30
gemma-4-E2B-it-qat-q4_0-unquantized5.1B 128K 11.8 GB4.0 GB 8GB GPUMac 16GB Commercial OK2026-04-29
gemma-4-E2B-it5.12B 128K 11.8 GB4.0 GB 8GB GPUMac 16GB Commercial OK2026-03-02
gemma-4-E4B-it8B 128K 18.2 GB6.0 GB 8GB GPUMac 16GB Commercial OK2026-03-02

04Embedding (2)

ModelParamsBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
embeddinggemma-2744M 2.1 GB1.0 GB 8GB GPUMac 16GB Commercial OK2026-09-14
embeddinggemma-300m303M 1.2 GB0.7 GB 8GB GPUMac 16GB Commercial with conditions2025-07-17

05Translation (3)

ModelParamsContextBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
translategemma-4b-it4.3B — 10.0 GB3.4 GB 8GB GPUMac 16GB Commercial with conditions2026-01-12
translategemma-12b-it12.2B — 27.3 GB8.6 GB 12GB GPUMac 16GB Commercial with conditions2026-01-12
translategemma-27b-it27.4B — 60.9 GB18.8 GB RTX 4090 24GBMac 36GB Commercial with conditions2026-01-12

06Safety / guard (2)

ModelParamsContextBF164-bitSmallest GPU (4-bit)Mac (4-bit)LicenceReleased
shieldgemma-2-4b-it4.3B — 10.0 GB3.4 GB 8GB GPUMac 16GB Commercial with conditions2025-03-04
shieldgemma-9b9.24B — 20.8 GB6.7 GB 8GB GPUMac 16GB Commercial with conditions2024-07-16

07Other families

Qwen · IBM Granite · NVIDIA Nemotron · Mistral / Magistral / Devstral · GLM / CogVideoX (Z.ai) · DeepSeek · Ai2 OLMo / Molmo · Ant Ling / Ring / Ming · Liquid LFM · MiniCPM / VoxCPM · Tencent Hunyuan / Hy · Phi · Llama · Meituan LongCat · Jina embeddings & rerankers · Wan · Kimi · Cohere Command / Aya · Xiaomi MiMo · FLUX · Microsoft VibeVoice / Fara / Harrier · InternLM / Intern-S · StepFun Step · MiniMax · Baidu ERNIE · NVIDIA Parakeet / Canary speech · Stable Diffusion / Stable Audio · Yi (01.AI) · ByteDance Seed · SmolLM / SmolVLM · BGE embeddings & rerankers · LTX Video · FunAudioLLM speech · gpt-oss · poolside Laguna · Whisper · Baichuan · Thinking Machines Inkling · xAI Grok · rednote dots

Totals include 10% + 0.5GB overhead and one 8K-token KV cache for text models. Estimates only. Method →

Data updated: Self-host open models(UTC+8)