gemma-4-31B-it

31.3B parameters, images and text in, text out, 256K tokens of context. 14 providers serve it from $0.10 per million input tokens.

Verified 2026-10-25 Rank #8 in LLMs · score 131.5 method v1 Sources 5 Downloads 30d 8.5M · #15 of 619 tracked

Changed this week: 1 new quantized build on the Hub (343 in all).2026-10-24

Can I run it?19GB at INT4Estimatefits a 24 GB cardmethod

What does it cost?$0.10/1M incheapest of 12CoreWeavefp42026-10-24

Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24

Do people use it?8.5Mdownloads, 30 days#15 of 619 we track · top 2%+46 likes, downloads −7% this week4,189 likesHugging Face2026-10-24

SVG reply: a pelican steals a bicycle from a supermarket — Draw a pelican stealing a bicycle from a supermarket. Reply with one complete SVG document only, no prose and no code fences. Use viewBox="0 0 200 150" and a transparent background. Use only these colours for every fill and stroke: #06e0c0 #ffc298 #cd85f9 #f1f2f2.

Finish the dad joke

as received● 1 line · 9 words

Why did the open-weights model cross the road?

To get to the other side of the license.

Finish this dad joke. Reply with the punchline only, on one line, without quotes: Why did the open-weights model cross the road?

A parasailer, in ASCII

as received● 12×16 fits 12×32

      ___
    /     \
   /       \
  /_________\
       |
       |
       o
      /|\
      / \
~~~~~~~~~~~~~~~
  ~  ~  ~  ~  ~
~~~~~~~~~~~~~~~~

Draw ASCII art of a parasailer over the sea. At most 12 lines and 32 columns. Reply with the art only: no code fences, no explanation.

Animate a bouncing ball

as received✕ 8 frames · 16×16

No playable frames in the reply.

Make a 16x16 pixel sprite animation of a ball bouncing, 8 frames. Reply with JSON only, no code fences: {"palette": {"a": "#rrggbb", ...}, "fps": <number>, "frames": [[16 strings of 16 characters], ...]}. At most 4 palette colours with single-letter keys; "." means transparent; exactly 8 frames.

The ringer: the same four playful tests for every model, replies shown as received and checked by code. A joke, not a benchmark; never ranked.3 of 4 pass · parasail-gemma-4-31b-it on parasail · run 2026-10-09 · temperature 0 · cost $0.001 · 2026-10 thing: supermarket · all models →

Worth reading

Picked by rule from 34 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 868 HN pts

Gemma 4 on iPhone

apps.apple.com 2026-04-05 868 HN pts · 234 comments about the gemma 4 family family

Latest · 59 days ago

Gemma 4 E2B inference in 700 lines of C

github.com 2026-08-27 18 HN pts 722 words about the gemma 4 family family

25 best of 34 tracked

The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

CodeShow HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Macgithub.com · 2026-07-29
ArticleGemma 4 12B: A unified, encoder-free multimodal modelblog.google · 2026-06-03 · official
ArticleRunning Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPUneomindlabs.com · 2026-07-15
CodeShow HN: Cactus Hybrid: We taught Gemma 4 to know when it's wronggithub.com · 2026-07-22 · 191 HN pts
ArticleGemma 4 QAT models: Optimizing compression for mobile and laptop efficiencyblog.google · 2026-06-05 · 406 HN pts · official
ArticleAccelerating Gemma 4: faster inference with multi-token prediction draftersblog.google · 2026-05-05 · 687 HN pts · official
ArticleGoogle releases Gemma 4 open modelsdeepmind.google · 2026-04-02 · 1,812 HN pts · official
ArticleFable 5 pushed Gemma 4 to 255 tok/s on WebGPUxcancel.com · 2026-06-18 · 48 HN pts
ArticleGoogle Gemma 4 Runs Natively on iPhone with Full Offline AI Inferencegizmoweek.com · 2026-04-15 · 303 HN pts
CodeApril 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac minigist.github.com · 2026-04-03
ArticleShow HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XPtokendelivery.ai · 2026-09-12 · 9 HN pts
ArticleI ran Gemma 4 as a local model in Codex CLIblog.danielvaughan.com · 2026-04-12 · 285 HN pts
ArticleRunning Gemma 4 locally with LM Studio's new headless CLI and Claude Codeai.georgeliu.com · 2026-04-05
ArticleGemma 4 on Cerebras - The Fastest Inference Is Now Multimodalcerebras.ai · 2026-06-30
CodeShow HN: Gemma 4 Multimodal Fine-Tuner for Apple Silicongithub.com · 2026-04-07
ArticleShow HN: Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)teamchong.github.io · 2026-04-19
CodeShow HN: Local privacy-first Microsoft Recall alternative with Gemma 4github.com · 2026-07-04
ArticleGemma 4: Byte for byte, the most capable open modelsblog.google · 2026-04-02 · official
ArticleA Visual Guide to Gemma 4 12Bnewsletter.maartengrootendorst.com · 2026-06-03
CodeShow HN: Local LLM code-generation with Gemma 4 e2B via JSON AST to Clojuregithub.com · 2026-05-19
ArticleShow HN: Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPUapeg.dev · 2026-06-30
ArticleShow HN: I made a Gemma 4 Mac app that names screenshots with local AIsnapname.app · 2026-05-31
ArticleGoogle's new Gemma 4 12B model is designed to run on any laptop with 16GB of RAMarstechnica.com · 2026-06-03

Show all 23

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

Prices for the same weights vary 8×

12 providers with a live price, from $0.10 to $0.75 per million input tokens; the median is $0.14. One dot per provider, cheapest in teal. 1 free-tier row not drawn.

#	Cheapest 5	In /1M	Out /1M	Context	Quant	Seen
1	CoreWeave	$0.10	$0.34	256K	fp4	Oct 24
2	Venice	$0.12	$0.36	250K	fp4	Oct 24
3	Chutes	$0.12	$0.37	128K	fp4	Oct 24
4	DeepInfra	$0.14	$0.51	256K	fp8	Oct 25
5	Crusoe	$0.14	$0.40	256K	bf16	Oct 24

All 14 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Widely served (14 API providers); heavily run locally (343 quantized builds); a big base for builders (730 fine-tunes, adapters and merges).

As a service14Chutes, CoreWeave, Crusoe, DeepInfra +10
On own hardware343quantized builds
Built on730327 fine-tunes · 363 adapters · 40 merges
In apps100+Hugging Face Spaces

Top Spaces running it: smolagents/hf-realtime-voice ♥ 584 · embedl/hfviewer ♥ 181 · victor/gemma-avatar ♥ 180 · multimodalart/krea2-lora-trainer ♥ 76

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 64 GB Mac (unified memory) at INT4

About 19 GB at INT4, 38 GB at FP8, 75 GB at BF16: weights only, so leave room for the KV cache.

Estimate params × bytes per param × 1.2 · how we estimate

Day-one receipts

Claim-matched tests, graded by code. Receipts, not a score.

Speed run● pass

{
  "providers": [
    {
      "provider": "parasail",
      "endpoint_model": "parasail-gemma-4-31b-it",
      "source_url": "https://api.parasail.io/v1/chat/completions",
      "ttft_ms_median": 1149,
      "ttft_any_ms_median": 1149,
      "tokens_per_s_median": 21.6,
      "runs": [
        {
          "ttft_s": 1.149,
          "ttft_any_s": 1.149,
          "tokens_per_s": 21.6,
          "latency_s": 7.6,
          "tokens_out": 164,
          "reasoning_chars": 0,
          "finish": "stop"
        },
        {
          "ttft_s": 0.691,
          "ttft_any_s": 0.691,
          "tokens_per_s": 27.8,
          "latency_s": 5.82,
          "tokens_out": 162,
          "reasoning_chars": 0,
          "finish": "stop"
        },
        {
          "ttft_s": 2.178,
          "ttft_any_s": 2.178,
          "tokens_per_s": 17.4,
          "latency_s": 9.19,
          "tokens_out": 160,
          "reasoning_chars": 0,
          "finish": "stop"
        }
      ]
    }
  ],
  "note": "Single client, wall clock; tok/s counts every output token including hidden reasoning."
}

Three-tool errand● pass

{
  "calls": [
    {
      "name": "get_weather",
      "args": {
        "city": "Lisbon",
        "date": "2026-10-14"
      }
    },
    {
      "name": "check_calendar",
      "args": {
        "date": "2026-10-14"
      }
    },
    {
      "name": "book_table",
      "args": {
        "date": "2026-10-14",
        "party_size": 2,
        "restaurant": "Cervejaria Ramiro",
        "time": "19:30"
      }
    }
  ],
  "checks": {
    "get_weather": true,
    "check_calendar": true,
    "book_table": true
  },
  "final": "The confirmation code for your table at Cervejaria Ramiro is TBL-4471.",
  "fixture": "receipts.errand.v1"
}

Messy addresses to JSON● pass

{
  "fields_matched": 80,
  "fields_total": 80,
  "cost_per_1000_usd": 0.0298,
  "parsed": true,
  "per_item": [
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    },
    {
      "street": true,
      "city": true,
      "postcode": true,
      "country": true
    }
  ],
  "sample": [
    {
      "street": "1600 Amphitheatre Parkway",
      "unit": null,
      "city": "Mountain View",
      "region": "CA",
      "postcode": "94043",
      "country": "US"
    },
    {
      "street": "10 Downing Street",
      "unit": null,
      "city": "London",
      "region": null,
      "postcode": "SW1A 2AA",
      "country": "GB"
    },
    {
      "street": "Rua Augusta 274",
      "unit": null,
      "city": "Lisboa",
      "region": null,
      "postcode": "1100-053",
      "country": "PT"
    }
  ],
  "fixture": "receipts.addresses.v1"
}

Needle heatmap● pass

{
  "lengths_k": [
    8,
    32,
    64
  ],
  "depths": [
    0.1,
    0.5,
    0.9
  ],
  "cells": [
    [
      1,
      1,
      1
    ],
    [
      1,
      1,
      1
    ],
    [
      1,
      1,
      1
    ]
  ],
  "detail": [
    {
      "k": 8,
      "depth": 0.1,
      "recalled": true,
      "reply": "amber-falcon-7731",
      "latency_s": 2.28,
      "finish": "stop"
    },
    {
      "k": 8,
      "depth": 0.5,
      "recalled": true,
      "reply": "cobalt-marten-7828",
      "latency_s": 1.63,
      "finish": "stop"
    },
    {
      "k": 8,
      "depth": 0.9,
      "recalled": true,
      "reply": "saffron-newt-7925",
      "latency_s": 2.68,
      "finish": "stop"
    },
    {
      "k": 32,
      "depth": 0.1,
      "recalled": true,
      "reply": "jade-lynx-8022",
      "latency_s": 9.72,
      "finish": "stop"
    },
    {
      "k": 32,
      "depth": 0.5,
      "recalled": true,
      "reply": "umber-kestrel-8119",
      "latency_s": 7.59,
      "finish": "stop"
    },
    {
      "k": 32,
      "depth": 0.9,
      "recalled": true,
      "reply": "coral-heron-8216",
      "latency_s": 7.13,
      "finish": "stop"
    },
    {
      "k": 64,
      "depth": 0.1,
      "recalled": true,
      "reply": "slate-badger-8313",
      "latency_s": 17.83,
      "finish": "stop"
    },
    {
      "k": 64,
      "depth": 0.5,
      "recalled": true,
      "reply": "ochre-otter-8410",
      "latency_s": 17.73,
      "finish": "stop"
    },
    {
      "k": 64,
      "depth": 0.9,
      "recalled": true,
      "reply": "teal-ibis-8507",
      "latency_s": 21.07,
      "finish": "stop"
    }
  ]
}

The story so far

Weights, every dated article, and repo commits on one rail.

10 dated articles and 13 repo commits so far

Teal is the weights (2026-03-11), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.

What people built on it

1,073 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

1,073 derivatives on the Hub, led by 363 adapters

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF unsloth/gemma-4-31B-it-GGUF for llama.cpp, Ollama, LM Studio 5.2M downloads MLX lmstudio-community/gemma-4-31B-it-MLX-8bit for Apple silicon 946.8k downloads NVFP4 nvidia/Gemma-4-31B-IT-NVFP4 for vLLM, SGLang 13.6M downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

16 rows

Benchmarks 16 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (16)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
author-reported AIME 2026 · Reason through hard problems	89.2% percent	Gemma 4 31B	model author	2026-10-07
author-reported BBEH · Reason through hard problems	74.4% percent	Gemma 4 31B	model author	2026-10-07
author-reported Codeforces · Write code with agents	2150 elo	Gemma 4 31B	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	84.3% percent	Gemma 4 31B	model author	2026-10-07
author-reported HLE with search	26.5% percent	Gemma 4 31B	model author	2026-10-07
author-reported Humanity's Last Exam · Reason through hard problems	19.5% percent	Gemma 4 31B	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	80.0% percent	Gemma 4 31B	model author	2026-10-07
author-reported MMLU-Pro · Reason through hard problems	85.2% percent	Gemma 4 31B	model author	2026-10-07
author-reported MMMLU · Work across languages	88.4% percent	Gemma 4 31B	model author	2026-10-07
author-reported MMMU-Pro · Understand images and video	76.9% percent	Gemma 4 31B	model author	2026-10-07
author-reported MRCR · Read long documents	66.4% percent	Gemma 4 31B	model author	2026-10-07
author-reported MathVision · Understand images and video	85.6% percent	Gemma 4 31B	model author	2026-10-07
author-reported MedXPertQA MM	61.3% percent	Gemma 4 31B	model author	2026-10-07
author-reported OmniDocBench · Extract text from documents	0.13 other	Gemma 4 31B	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	76.9% percent	Gemma 4 31B	model author	2026-10-07
author-reported mmmu_pro_vision (MMMU/MMMU_Pro)	76.9	—	Model Card	2026-05-12

What the card says

1 cited sentence

Summary generated, cited

Gemma 4 31B-it is an instruction-tuned, multimodal model with 31.3B parameters.^[1]^[2]

parameters human
“instruction-tuned”

Generated 2026-10-07 14:59 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

8 use cases

What it is good at 8 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

21 fields

Listed in LLMs

Facts

gemma-4-31B-it has 31.3B parameters. (safetensors metadata, observed 2026-10-24)
Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
Created 2026-03-11, last modified 2026-07-20. (HF API, observed 2026-10-24)

Pipeline tag	image-text-to-text
Library	transformers
Architectures	Gemma4ForConditionalGeneration
Model type	gemma4
Dtype	bfloat16
Params by dtype	BF16: 31,273,088,876
Size band	L
MoE	no
Context length	262,144 (config.json max_position_embeddings)
Papers	arXiv:2607.02770
License	Apache 2.0 (apache-2.0)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Gemma4ForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Gemma4ForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for gemma-4-31B-it: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	75 GB · Single high-end GPU (<80 GB)
FP8 (1 byte/param)	38 GB · Single high-end GPU (<80 GB)
INT4 (0.5 bytes/param)	19 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Fine-tuned from google/gemma-4-31B (same organization) Family: google/gemma-4 1,073 derivatives on the Hub

Base model: google/gemma-4-31B

quantized: 343finetune: 327adapter: 363merge: 40

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
nvidia/Gemma-4-31B-IT-NVFP4	quantized	NVFP4	20.9B	L	other	nvidia	586	13.6M
RedHatAI/gemma-4-31B-it-FP8-block	quantized	-FP8	31.3B	L	apache-2.0	RedHatAI	52	10.7M
unsloth/gemma-4-31B-it-GGUF	quantized	GGUF	—	—	apache-2.0	unsloth	651	5.2M
cyankiwi/gemma-4-31B-it-AWQ-4bit	quantized	AWQ	31.3B	L	apache-2.0	cyankiwi	55	4.6M
LilaRest/gemma-4-31B-it-NVFP4-turbo	quantized	NVFP4	30.7B	L	apache-2.0	LilaRest	307	2.3M
QuantTrio/gemma-4-31B-it-AWQ	quantized	AWQ	31.3B	L	apache-2.0	QuantTrio	14	2.2M
RedHatAI/gemma-4-31B-it-NVFP4	quantized	NVFP4	32.7B	L	apache-2.0	RedHatAI	59	2.1M
wangzhang/gemma-4-31B-it-abliterated	finetune	—	31.3B	L	apache-2.0	wangzhang	39	2.1M
unsloth/gemma-4-31B-it-unsloth-bnb-4bit	quantized	-bnb	31.3B	L	apache-2.0	unsloth	25	2M
RedHatAI/gemma-4-31B-it-FP8-dynamic	quantized	-FP8	31.3B	L	apache-2.0	RedHatAI	24	1.7M
paperscarecrow/Gemma-4-31B-it-abliterated	quantized	GGUF	—	—	apache-2.0	paperscarecrow	124	1.5M
lmstudio-community/gemma-4-31B-it-GGUF	quantized	GGUF	—	—	apache-2.0	lmstudio-community	29	1.3M
lmstudio-community/gemma-4-31B-it-MLX-8bit	quantized	MLX	31.3B	L	apache-2.0	lmstudio-community	12	946.8k
lmstudio-community/gemma-4-31B-it-MLX-4bit	quantized	MLX	31.3B	L	apache-2.0	lmstudio-community	2	824.9k
lmstudio-community/gemma-4-31B-it-MLX-6bit	quantized	MLX	31.3B	L	apache-2.0	lmstudio-community	2	779.7k
bartowski/google_gemma-4-31B-it-GGUF	quantized	GGUF	—	—	apache-2.0	bartowski	96	685.4k
HauhauCS/Gemma4-31B-QAT-Uncensored-HauhauCS-Balanced-MTP	quantized	GGUF	—	—	gemma	HauhauCS	194	661.3k
Intel/gemma-4-31B-it-int4-AutoRound	quantized	int4	5.90B	M	—	Intel	55	559.8k
lmstudio-community/gemma-4-31B-it-MLX-5bit	quantized	MLX	31.3B	L	apache-2.0	lmstudio-community	2	498.6k
google/gemma-4-31B-it-qat-q4_0-unquantized	finetune	q4_0	31.3B	L	apache-2.0	google	54	367.5k
cyankiwi/gemma-4-31B-it-AWQ-8bit	quantized	AWQ	31.3B	L	apache-2.0	cyankiwi	19	343.7k
TrevorJS/gemma-4-31B-it-uncensored	finetune	—	32.7B	L	apache-2.0	TrevorJS	34	308.9k
ebircak/gemma-4-31B-it-4bit-W4A16-AWQ	quantized	-4bit	31.7B	L	apache-2.0	ebircak	3	297.5k
mlx-community/gemma-4-31b-it-4bit	quantized	-4bit	31.3B	L	apache-2.0	mlx-community	50	283.4k
ggml-org/gemma-4-31B-it-GGUF	quantized	GGUF	—	—	apache-2.0	ggml-org	47	235.2k
unsloth/gemma-4-31B-it	finetune	—	32.7B	L	apache-2.0	unsloth	22	212.7k
llmfan46/gemma-4-31B-it-uncensored-heretic	finetune	—	31.3B	L	apache-2.0	llmfan46	162	105.2k
unsloth/gemma-4-31B-it-NVFP4	quantized	NVFP4	21.6B	L	apache-2.0	unsloth	32	105.3k
aifeifei798/Gemma-4-31B-Cognitive-Unshackled	finetune	—	31.3B	L	apache-2.0	aifeifei798	29	94.7k
huihui-ai/Huihui-gemma-4-31B-it-abliterated-v2	finetune	—	32.7B	L	apache-2.0	huihui-ai	41	74.5k
TheDrummer/Artemis-31B-v1-GGUF	quantized	GGUF	—	—	—	TheDrummer	29	67.6k
LiconStudio/Gemma-4-31B-it-abliterated-GGUF	quantized	GGUF	—	—	apache-2.0	LiconStudio	15	66.5k
coder3101/gemma-4-31B-it-heretic	finetune	—	31.3B	L	apache-2.0	coder3101	91	57.3k
QuantTrio/gemma-4-31B-it-AWQ-6Bit	quantized	AWQ	31.3B	L	apache-2.0	QuantTrio	10	58.3k
prithivMLmods/gemma-4-31B-it-FP8	quantized	-FP8	32.7B	L	apache-2.0	prithivMLmods	4	55.3k
Momix-44/gemma-4-31B-it-heretic-v2	quantized	GGUF	31.3B	L	apache-2.0	Momix-44	7	54.3k
mlx-community/gemma-4-31b-it-8bit	quantized	-8bit	31.3B	L	apache-2.0	mlx-community	20	46.3k
DavidAU/gemma-4-31B-it-The-DECKARD-HERETIC-UNCENSORED-Thinking	finetune	—	31.3B	L	apache-2.0	DavidAU	94	38.6k
ebircak/gemma-4-31B-it-4bit-W4A16-GPTQ	quantized	-4bit	31.7B	L	apache-2.0	ebircak	3	40.4k
slashreboot/athena-class-model-a	finetune	GGUF	—	—	apache-2.0	slashreboot	7	36.7k

Repository changelog

13 commits

changed in the last 7 daysOther: 2026-10-24all changes

Repository changelog 13 commits

Date	Commit	SHA
2026-07-20	Add response_template to tokenizer_config.json (#130)	842da37
2026-07-15	Update README (#133)	b9ea41a
2026-07-15	fix: chat template — null handling, reasoning preservation, turn-tag balance, input validation (#118)	68abe48
2026-07-10	Add Technical Report	518276f
2026-06-03	Update README.md	3548789
2026-06-02	Update README.md	ec81324
2026-05-27	Add MMMU-Pro evaluation result (#107)	fb9ae26
2026-05-18	fix(chat_template): Emit multimodal placeholders in tool response content-parts (#94)	fcf2302
2026-05-07	Add proper metadata (#101)	ba74f5b
2026-04-28	fix(chat_template): update SI and tool call handling (#86)	145dc25
2026-04-10	Update schema (#59)	439edf5
2026-04-10	fix: prefix-preserving tool calls + dialog compliance (#56)	e51e7dc
2026-04-02	Preparing for release!	419b2ef

All providers

14 providers · 14 endpoints

Where to run 14 providers · 14 endpoints

Hosted endpoints observed for gemma-4-31B-it, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.00/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
Google AI Studio openrouter	google/gemma-4-31b-it	256K	unknown	$0.00	$0.00	—	—	yes	live	2026-10-24
CoreWeave openrouter	google/gemma-4-31b-it	256K	fp4	$0.10	$0.34	—	—	yes	live	2026-10-24
Venice openrouter	google/gemma-4-31b-it	250K	fp4	$0.12	$0.36	—	—	yes	live	2026-10-24
Chutes openrouter	google/gemma-4-31b-it	128K	fp4	$0.12	$0.37	—	—	yes	live	2026-10-24
DeepInfra hf, deepinfra, openrouter	google/gemma-4-31B-it	256K	fp8	$0.14	$0.51	36.8	301 ms	yes	live	2026-10-25
Novita hf, openrouter	google/gemma-4-31b-it	256K	bf16	$0.14	$0.40	56	674 ms	yes	live	2026-10-25
Crusoe openrouter	google/gemma-4-31b-it	256K	bf16	$0.14	$0.40	—	—	yes	live	2026-10-24
Friendli openrouter	google/gemma-4-31b-it	256K	unknown	$0.14	$0.40	—	—	yes	live	2026-10-24
Parasail parasail, openrouter	parasail-gemma-4-31b-it	256K	fp8	$0.15	$0.40	—	—	yes	serverless	2026-10-24
Io Net openrouter	google/gemma-4-31b-it	256K	unknown	$0.36	$1.09	—	—	yes	live	2026-10-24
SambaNova openrouter	google/gemma-4-31b-it	128K	unknown	$0.38	$1.15	—	—	yes	live	2026-10-24
ModelRun openrouter	google/gemma-4-31b-it	256K	fp4	$0.75	$1.00	—	—	yes	live	2026-10-24
SiliconFlow openrouter	google/gemma-4-31b-it	256K	fp8	$0.75	$1.00	—	—	yes	live	2026-10-24
Featherless hf	google/gemma-4-31B-it	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

34 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official3

Gemma 4 QAT models: Optimizing compression for mobile and laptop efficiency

blog.google2026-06-05officialauthorityHN front pageabout the gemma 4 familyHN 406 pts · 130 comments714 words

Google released new Gemma 4 checkpoints optimized with Quantization-Aware Training to improve efficiency on edge devices and consumer GPUs. The E2B model achieves a 1GB memory footprint using a mobile-specialized schema that applies targeted 2-bit quantization to token generation layers. Developers can access GGUF formats ready for llama.cpp and MTP QAT checkpoints to preserve inference speedups while running locally.written, reviewed, cited

The mobile-specialized schema pre-calculates activation scaling settings during training to reduce workload on mobile chips.
Targeted 2-bit quantization compresses token generation parts while keeping core reasoning layers at higher precision.
Deploying only needed modalities allows the Gemma 4 E2B text-only model to require less than 1 GB of memory.

Accelerating Gemma 4: faster inference with multi-token prediction drafters

blog.google2026-05-05officialauthorityHN front pageabout the gemma 4 familyHN 687 pts · 330 comments796 words

Google released Multi-Token Prediction (MTP) drafters for the Gemma 4 family under the same open-source Apache 2.0 license as Gemma 4. These drafters deliver up to a 3x speedup for models like the 26B MoE and 31B Dense versions without quality degradation. Benchmarks on hardware such as the NVIDIA RTX PRO 6000 show that increasing batch sizes to 4 to 8 unlocks up to a ~2.2x speedup locally.written, reviewed, cited

The MTP drafter shares the target model's KV cache to avoid recalculating context the larger model has already processed.
For E2B and E4B edge models, an efficient clustering technique in the embedder accelerates generation where logit calculation is a bottleneck.
The 26B mixture-of-experts model presents unique routing challenges at a batch size of 1 on Apple Silicon.

Google releases Gemma 4 open models

deepmind.google2026-04-02officialauthorityHN front pageabout the gemma 4 familyHN 1,812 pts · 474 comments253 words

Analysis7

Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong

github.com2026-07-22authorityHN front pageabout the gemma 4 familyHN 191 pts · 44 comments714 words

Cactus Hybrid introduces on-device models that score every answer with a confidence between 0 and 1 to decide when to hand off to a bigger model. The Gemma 4 E2B Hybrid model matches Gemini 3.1 Flash-Lite performance by routing only 15–55% of queries to the cloud while running the rest locally. It is MIT-licensed and achieves a 0.770 AUROC on MMLU, with a probe trained on zero audio data yet reaching 0.79–0.88 AUROC on four audio benchmarks.written, reviewed, cited

Weights on the meta device crash the confidence read because probes score generations outside the module forward path.
The probe is a patch you compile into the llama.cpp engine rather than a standalone library component.
Transformers version 5.14 and above cause a segfault on this specific checkpoint.

Gemma 4 E2B inference in 700 lines of C

github.com2026-08-27authorityabout the gemma 4 familyHN 18 pts · 0 comments722 words

Fable 5 pushed Gemma 4 to 255 tok/s on WebGPU

xcancel.com2026-06-18about the gemma 4 familyHN 48 pts · 22 comments

Gemma 4 on iPhone

apps.apple.com2026-04-05HN front pageabout the gemma 4 familyHN 868 pts · 234 comments

Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference

gizmoweek.com2026-04-15HN front pageabout the gemma 4 familyHN 303 pts · 187 comments364 words

Show HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XP

tokendelivery.ai2026-09-12about the gemma 4 familyHN 9 pts · 1 comments

I ran Gemma 4 as a local model in Codex CLI

blog.danielvaughan.com2026-04-12HN front pageabout the gemma 4 familyHN 285 pts · 116 comments

All coverage tracked for this model (34) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
Show HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XP	tokendelivery.ai	2026-09-12	2026-10-07	hn
Gemma 4 E2B inference in 700 lines of C	github.com	2026-08-27	2026-10-07	hn
Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac	github.com	2026-07-29	2026-10-07	hn
Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong	github.com	2026-07-22	2026-10-07	hn
Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU	neomindlabs.com	2026-07-15	2026-10-07	hn
Show HN: Local privacy-first Microsoft Recall alternative with Gemma 4	github.com	2026-07-04	2026-10-07	hn
Show HN: Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPU	apeg.dev	2026-06-30	2026-10-07	hn
Gemma 4 on Cerebras - The Fastest Inference Is Now Multimodal	cerebras.ai	2026-06-30	2026-10-07	hn
Fable 5 pushed Gemma 4 to 255 tok/s on WebGPU	xcancel.com	2026-06-18	2026-10-07	hn
Gemma 4 QAT models: Optimizing compression for mobile and laptop efficiency	blog.google official	2026-06-05	2026-10-07	hn
Google's new Gemma 4 12B model is designed to run on any laptop with 16GB of RAM	arstechnica.com	2026-06-03	2026-10-07	hn
A Visual Guide to Gemma 4 12B	newsletter.maartengrootendorst.com	2026-06-03	2026-10-07	hn
Gemma 4 12B: A unified, encoder-free multimodal model	blog.google official	2026-06-03	2026-10-07	hn
Show HN: I made a Gemma 4 Mac app that names screenshots with local AI	snapname.app	2026-05-31	2026-10-07	hn
Show HN: Local LLM code-generation with Gemma 4 e2B via JSON AST to Clojure	github.com	2026-05-19	2026-10-07	hn
Accelerating Gemma 4: faster inference with multi-token prediction drafters	blog.google official	2026-05-05	2026-10-07	hn
Show HN: Prompt-to-Excalidraw demo with Gemma 4 E2B in the browser (3.1GB)	teamchong.github.io	2026-04-19	2026-10-07	hn
Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference	gizmoweek.com	2026-04-15	2026-10-07	hn
Show HN: I benchmarked Gemma 4 E2B – the 2B model beat the 12B on multi-turn	aiexplr.com	2026-04-13	2026-10-07	hn
I ran Gemma 4 as a local model in Codex CLI	blog.danielvaughan.com	2026-04-12	2026-10-07	hn
A Visual Guide to Gemma 4	newsletter.maartengrootendorst.com	2026-04-09	2026-10-07	hn
Show HN: Gemma 4 Multimodal Fine-Tuner for Apple Silicon	github.com	2026-04-07	2026-10-07	hn
Gemma 4 on iPhone	apps.apple.com	2026-04-05	2026-10-07	hn
Running Gemma 4 locally with LM Studio's new headless CLI and Claude Code	ai.georgeliu.com	2026-04-05	2026-10-07	hn
Gemma 4: Google DeepMind's Most Intelligent Open Models \| AI \| Tobias Weiss	tobias-weiss.org	2026-04-04	2026-10-07	exa
Gemma 4 Review: Apache-2.0 Open AI, 89.2% on AIME 2026 \| Nerd Level Tech	nerdleveltech.com	2026-04-04	2026-10-07	exa
April 2026 TLDR Setup for Ollama and Gemma 4 26B on a Mac mini	gist.github.com	2026-04-03	2026-10-07	hn
Google releases Gemma 4 open models	deepmind.google official	2026-04-02	2026-10-07	hn
Gemma 4: Byte for byte, the most capable open models	blog.google official	2026-04-02	2026-10-07	hn
Gemma 4 Architecture and Benchmark Notes \| Sebastian Raschka, PhD	sebastianraschka.com	2026-04-02	2026-10-07	exa
Gemma 4 31B: benchmarks, API pricing & context window — ModelCap #20	modelcap.ai	2026-04-02	2026-10-07	exa
Gemma 4 31B: 262K Context & GPQA Diamond Score \| HokAI	hokai.io	2026-04-02	2026-10-07	exa
Gemma 4 31B IT: benchmarks, pricing & specs	themodelbeat.com	2026-04-02	2026-10-07	exa
Google Gemma 4 Deep Dive: Architecture, MoE & Benchmarks	qubridai.hashnode.dev	2026-04-02	2026-10-07	exa

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

Gemma 4 Technical Report this model

Gemma Team, Sherif El Abd, Vaibhav Aggarwal +320 · 2026-07-02 · PDF · HF paper page (85 upvotes)

Similar models

3 models

Similar models

Qwen3.8-27B · 27.8B
same size band (L), same task (image-text-to-text), more providers, permissive license
Qwen3.6-27B · 27.8B
same size band (L), same task (image-text-to-text), fewer providers, permissive license
Muse-Glimmer-30B · 29.8B
same size band (L), same task (image-text-to-text), fewer providers, permissive license

Sources

5 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:31 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 14:59 UTC · Last verified 2026-10-25

Follow gemma-4-31B-it

Watch this model

/feeds/models/google--gemma-4-31B-it.xml(RSS 2.0) · every change to gemma-4-31B-it
/feeds/families/google--gemma-4.xml(RSS 2.0) · every change across the gemma-4 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.

The ringer: four playful tests

A pelican steals a bicycle from a supermarket

Finish the dad joke

A parasailer, in ASCII

Animate a bouncing ball

Worth reading

Gemma 4 on iPhone

Gemma 4 E2B inference in 700 lines of C

25 best of 34 tracked

Where to run it

Prices for the same weights vary 8×

Where it's used

Will it fit?

Fits a 64 GB Mac (unified memory) at INT4

Day-one receipts

The story so far

10 dated articles and 13 repo commits so far

What people built on it

1,073 derivatives on the Hub, led by 363 adapters

Reference

Benchmarks

Benchmarks 16 rows

What the card says

Summary generated, cited

Use-case evidence

What it is good at 8 use cases

Facts

Facts

Servability

Servability

Hardware fit

Hardware fit ESTIMATE

All variants

Lineage

Top variants (40 shown, ranked by downloads)

Repository changelog

Repository changelog 13 commits

All providers

Where to run 14 providers · 14 endpoints

Articles and coverage

Official3

Analysis7

Papers

Papers 1

Similar models

Similar models

Sources

Sources

Follow gemma-4-31B-it

Watch this model