Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › GLM-4.6V family

GLM-4.6V-Flash

Hugging Face ↗ Follow changes

zai-org MIT out 11 months ago

10.3B parameters, images and text in, text out, 128K tokens of context. 2 providers serve it from $0.30 per million input tokens.

Verified 2026-10-25 Rank #224 in LLMs · score 87.6 method v1 Sources 3 Downloads 30d 81k · #337 of 619 tracked

Can I run it?6.2GB at INT4Estimatefits a 16 GB cardmethod
What does it cost?$0.30/1M incheapest of 1Novita2026-10-25
Can I use it?MITPermissivelicence file2026-10-24
Do people use it?81kdownloads, 30 days#337 of 619 we track · below the median+1 likes, downloads −21% this week631 likesHugging Face2026-10-24

Worth reading

Picked by rule from 10 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Latest · 34 days ago

GLM-4.6V review: open vision model, now overtaken by GLM-5.3-Flash | Recaply

GLM-4.6V is a vision-language model family from Z.ai that reads video, images, and text with a 128K token context.

  • The 9B GLM-4.6V-Flash model runs locally on a 12 GB gaming GPU or a 24 GB MacBook Pro.
  • GLM-4.6V-FlashX costs $0.04 per million input tokens and $0.4 per million output tokens on the API.
recaply.co 2026-09-22 1.5k words about the GLM 4.6V family family

10 best of 10 tracked

The 2 picks above, plus these 8. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. ArticleGLM 4.6V - API Pricing & Benchmarks | OpenRouteropenrouter.ai · undated
  2. ArticleGLM-4.6Vz.ai · 2025-12-08 · 8 HN pts · official
  3. ArticleGLM-4.6V Arrives: Seamless Integration of Visual Perception and Action Execution | Communeifycommuneify.com · 2026-07-17
  4. ArticleGLM-4.6V — specs, benchmarks & price | AI/TLDRai-tldr.dev · undated
  5. ArticleGLM-4.6V-Flash vs GLM-4.7-Flash — Pricing, Benchmarks & Performance Comparedanotherwrapper.com · undated
  6. ArticleGLM-4.6V Review: Free 9B Flash Beats Qwen2-VL? [Benchmarks]binaryverseai.com · 2025-12-09
  7. ArticleGLM 4.6V Review: Best Open-Source Multimodal Modelblog.siray.ai · 2025-12-09
  8. ArticleGLM-4.6V: Specs & Benchmarks | LLMLearnerllmlearner.com · 2025-12-08

Where to run it

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

#Cheapest 2In /1MOut /1MContextQuantSeen
1 Novita $0.30 $0.90 128K — Oct 25
2 Z.ai — — — — Oct 25

All 2 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.

Served by 2 API providers; 49 quantized builds for local use.

  • As a service2Novita, Z.ai
  • On own hardware49quantized builds
  • Built on1814 fine-tunes · 2 adapters · 2 merges
  • In apps52Hugging Face Spaces

Top Spaces running it: hadadxyz/ai ♥ 819 · anycoderapps/GLM-4.6V-Flash ♥ 8 · Prabudh07/GLM-4.6V-Flash ♥ 1 · Open-Hat-Lab/GLM-4.6V-Flash ♥ 1

Will it fit?

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

Fits a 16 GB card at INT4

About 6.2 GB at INT4, 12 GB at FP8, 25 GB at BF16: weights only, so leave room for the KV cache.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 6.2 GB of weights (estimate; KV cache and activations extra)INT46.2 GBFP8: about 12 GB of weights (estimate; KV cache and activations extra)FP812 GBBF16: about 25 GB of weights (estimate; KV cache and activations extra)BF1625 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 6.2 GB of weights (estimate; KV cache and activations extra)INT46.2 GBFP8: about 12 GB of weights (estimate; KV cache and activations extra)FP812 GBBF16: about 25 GB of weights (estimate; KV cache and activations extra)BF1625 GB

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

The story so far

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

Most coverage came in the first two weeks

Teal is the weights (2025-12-07), amber triangles are 6 dated articles (each links to it), grey ticks are repo commits.

JanFebMarAprMayJunJulAugSepOct2025-12-09: 2 repo commits2025-12-08: 5 repo commits2025-12-07: 2 repo commitsGLM-4.6V · 2025-12-08GLM-4.6V review: open vision model, now overtaken by GLM-5.3-Flash | Recaply · 2026-09-22GLM-4.6V Arrives: Seamless Integration of Visual Perception and Action Execution | Communeify · 2026-07-17GLM-4.6V: Specs & Benchmarks | LLMLearner · 2025-12-08GLM-4.6V Review: Free 9B Flash Beats Qwen2-VL? [Benchmarks] · 2025-12-09GLM 4.6V Review: Best Open-Source Multimodal Model · 2025-12-09Weights 2025-12-07 · 2025-12-07Weights 2025-12-07now · 2 providers
JanFebMarAprMayJunJulAugSepOct2025-12-09: 2 repo commits2025-12-08: 5 repo commits2025-12-07: 2 repo commitsGLM-4.6V · 2025-12-08GLM-4.6V review: open vision model, now overtaken by GLM-5.3-Flash | Recaply · 2026-09-22GLM-4.6V Arrives: Seamless Integration of Visual Perception and Action Execution | Communeify · 2026-07-17GLM-4.6V: Specs & Benchmarks | LLMLearner · 2025-12-08GLM-4.6V Review: Free 9B Flash Beats Qwen2-VL? [Benchmarks] · 2025-12-09GLM 4.6V Review: Best Open-Source Multimodal Model · 2025-12-09Weights 2025-12-07 · 2025-12-07Weights 2025-12-07now · 2 providers

What people built on it

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

67 derivatives on the Hub, most of them quantized

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

49 quantized (Hugging Face model tree)14 fine-tunes (Hugging Face model tree)2 merges (Hugging Face model tree)2 adapters (Hugging Face model tree)49 quantized14 fine-tunes2 merges2 adapters
49 quantized (Hugging Face model tree)14 fine-tunes (Hugging Face model tree)2 merges (Hugging Face model tree)2 adapters (Hugging Face model tree)49 quantized14 fine-tunes2 merges2 adapters
GGUF MaziyarPanahi/GLM-4.6V-Flash-GGUF for llama.cpp, Ollama, LM Studio 713k downloads MLX lmstudio-community/GLM-4.6V-Flash-MLX-4bit for Apple silicon 1.9M downloads AWQ cyankiwi/GLM-4.6V-Flash-AWQ-4bit for vLLM, SGLang 27.2k downloads

All variants →

Reference

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

What the card says

1 cited sentence

Summary generated, cited

GLM-4.6V-Flash is a 10.3B parameter image-text-to-text model from the GLM-V family.[1][2][3]

  1. parameters human
  2. pipeline tag
  3. “part of the GLM-V family”

Generated 2026-10-07 15:05 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

7 use cases

What it is good at 7 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

22 fields

Listed in LLMs

Facts

Pipeline tagimage-text-to-text
Librarytransformers
ArchitecturesGlm4vForConditionalGeneration
Model typeglm4v
Dtypebfloat16
Params by dtypeBF16: 10,292,777,472
Size bandL
MoEno
Context length131,072 (config.json max_position_embeddings)
Languageszh, en
PapersarXiv:2507.01006
LicenseMIT (mit)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Glm4vForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Glm4vForConditionalGeneration · 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 GLM-4.6V-Flash: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)25 GB · Single high-end GPU (<80 GB)
FP8 (1 byte/param)12 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)6.2 GB · Single GPU (<24 GB)

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

All variants

40 variants

Lineage

Original model Family: zai-org/GLM-4.6V 67 derivatives on the Hub

quantized: 49finetune: 14merge: 2adapter: 2

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
lmstudio-community/GLM-4.6V-Flash-MLX-4bit quantized MLX — — mit lmstudio-community 3 1.9M
lmstudio-community/GLM-4.6V-Flash-MLX-8bit quantized MLX — — mit lmstudio-community 1 1.9M
lmstudio-community/GLM-4.6V-Flash-MLX-6bit quantized MLX — — mit lmstudio-community 0 1.8M
MaziyarPanahi/GLM-4.6V-Flash-GGUF quantized GGUF — — — MaziyarPanahi 7 713k
lmstudio-community/GLM-4.6V-Flash-GGUF quantized GGUF — — mit lmstudio-community 5 558.2k
unsloth/GLM-4.6V-Flash-GGUF quantized GGUF — — mit unsloth 126 373.6k
bartowski/zai-org_GLM-4.6V-Flash-GGUF quantized GGUF — — mit bartowski 18 61k
ggml-org/GLM-4.6V-Flash-GGUF quantized GGUF — — — ggml-org 24 40.7k
cyankiwi/GLM-4.6V-Flash-AWQ-4bit quantized AWQ 3.41B M mit cyankiwi 3 27.2k
cyankiwi/GLM-4.6V-Flash-AWQ-8bit quantized AWQ 4.43B M mit cyankiwi 0 23.7k
kalle07/GLM-4.6V-Flash-heretic_R18_K001-gguf quantized gguf — — mit kalle07 2 15.4k
mradermacher/GLM-4.6V-Flash-i1-GGUF quantized GGUF — — mit mradermacher 0 13.7k
mradermacher/GLM-4.6V-Flash-GGUF quantized GGUF — — mit mradermacher 1 12.3k
prithivMLmods/GLM-4.6V-Flash-GGUF quantized GGUF — — mit prithivMLmods 1 11.7k
AliBilge/Huihui-GLM-4.6V-Flash-abliterated quantized GGUF — — apache-2.0 AliBilge 1 8.1k
eaddario/GLM-4.6V-Flash-GGUF quantized GGUF — — mit eaddario 4 6.4k
unsloth/GLM-4.6V-Flash finetune — 10.3B L mit unsloth 5 5.8k
AaryanK/GLM-4.6V-Flash-GGUF quantized GGUF — — mit AaryanK 1 5.2k
Ex0bit/Elbaz-GLM-4.6V-Flash-PRISM quantized GGUF — — apache-2.0 Ex0bit 6 4.9k
Firworks/GLM-4.6V-Flash-nvfp4 quantized nvfp4 6.35B M mit Firworks 1 4.5k
NikolayKozloff/GLM-4.6V-Flash-Q8_0-GGUF quantized Q8_0 — — mit NikolayKozloff 3 3.8k
sartajbhuvaji/GLM-4.6-Flash-text finetune GGUF 9.40B L mit sartajbhuvaji 2 3.3k
neody/glm-4.6v-flash-gguf-text-only quantized gguf — — mit neody 0 2.8k
Edge-Quant/GLM-4.6V-Flash-Q4_K_M-GGUF quantized GGUF — — mit Edge-Quant 1 2.2k
AbteeXAILab/lumynax-multimodal-glm46v-flash quantized GGUF — — mit AbteeXAILab 0 2.3k
firedancer/GLM-4.6V-Flash-GGUF quantized GGUF — — — firedancer 0 1.7k
NikolayKozloff/GLM-4.6V-Flash-Q4_K_M-GGUF quantized GGUF — — mit NikolayKozloff 2 1.5k
mlx-community/GLM-4.6V-Flash-mxfp4 quantized mxfp4 10.3B L mit mlx-community 3 1.4k
DevQuasar/zai-org.GLM-4.6V-Flash-GGUF quantized GGUF — — — DevQuasar 0 1.2k
FreedomAISVR/GLM-4.6V-Flash-NVFP4-GGUF quantized NVFP4 — — mit FreedomAISVR 0 1.2k
FreedomAISVR/GLM-4.6V-Flash-MXFP4-GGUF quantized MXFP4 — — mit FreedomAISVR 2 972
alexgusevski/Huihui-GLM-4.6V-Flash-abliterated-q8-mlx quantized mlx 10.3B L mit alexgusevski 0 865
steampunque/GLM-4.6V-Flash-MP-GGUF quantized GGUF — — mit steampunque 0 736
alexgusevski/Huihui-GLM-4.6V-Flash-abliterated-q4-mlx quantized mlx 10.3B L mit alexgusevski 0 649
Mitchins/GLM-4.6V-Flash-FP8-Block128 quantized -FP8 10.3B L mit Mitchins 0 627
alexgusevski/Huihui-GLM-4.6V-Flash-abliterated-q2-mlx quantized mlx 10.3B L mit alexgusevski 0 617
NikolayKozloff/GLM-4.6V-Flash-Q5_K_M-GGUF quantized GGUF — — mit NikolayKozloff 2 481
jamiefutch/GLM-4.6V-Flash-MXFP4-GGUF quantized MXFP4 — — mit jamiefutch 0 541
alexgusevski/Huihui-GLM-4.6V-Flash-abliterated-q6-mlx quantized mlx 10.3B L mit alexgusevski 0 534
Mitchins/GLM-4.6V-Flash-NVFP4-BF16Vision quantized NVFP4 6.72B M mit Mitchins 0 437

Repository changelog

9 commits

Repository changelog 9 commits

DateCommitSHA
2025-12-09 Glm4vMoeForConditionalGeneration change to Glm4vForConditionalGeneration 411bb4d
2025-12-09 Update README.md (#5) f14ac1a
2025-12-08 api link 16cbd8b
2025-12-08 update 310c17d
2025-12-08 update 9c31aa1
2025-12-08 update 1f36507
2025-12-08 update 929496d
2025-12-07 Add files using upload-large-folder tool 838242d
2025-12-07 initial commit e465319

All providers

2 providers · 2 endpoints

Where to run 2 providers · 2 endpoints

Hosted endpoints observed for GLM-4.6V-Flash, 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: $0.30/M.

ProviderModel refContextQuant$/M in$/M outtok/sTTFTToolsStatusObserved
Novita
hf
 zai-org/glm-4.6v 128K — $0.30 $0.90 42.4 1228 ms yes live 2026-10-25
Z.ai
hf
 glm-4.6v-flash — — — — 66.3 4192 ms yes live 2026-10-25

Articles and coverage

10 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.

Official1

GLM-4.6V
z.ai2025-12-08officialauthorityabout the GLM 4.6V familyHN 8 pts · 0 comments1,383 words
Z.ai officially introduces the GLM-4.6V series, which integrates native Function Calling capabilities for the first time. The release includes GLM-4.6V (106B) for cloud scenarios and GLM-4.6V-Flash (9B) for local deployment, both scaling context to 128k tokens. The model achieves SOTA performance on benchmarks including MMBench, MathVista, and OCRBench among open-source models of comparable scale.written, reviewed, cited
  • GLM-4.6V accepts images and documents directly as tool parameters without converting them to text first.
  • The model processes approximately 150 pages of complex documents or a one-hour-long video in a single inference pass.
  • The team extends the Model Context Protocol to use URLs for identifying multimodal content passed to and from tools.

Analysis9

GLM 4.6V - API Pricing & Benchmarks | OpenRouter
openrouter.aiauthorityabout the GLM 4.6V family266 words
GLM-4.6V is a large multimodal model released on December 8, 2025, designed for high-fidelity visual understanding and long-context reasoning. It supports up to 131K tokens and costs $0.30 / $0.90 per 1M tokens for input and output respectively. The model is not currently available through EU in-region routing.written, reviewed, cited
GLM-4.6V review: open vision model, now overtaken by GLM-5.3-Flash | Recaply
recaply.co2026-09-22about the GLM 4.6V family1,514 words
GLM-4.6V is a vision-language model family from Z.ai that reads video, images, and text with a 128K token context. The full 106B model costs $0.3 / $0.9 per million tokens on the API, while the 9B GLM-4.6V-Flash variant is free and available under the MIT license for local use. Despite its capabilities, the model suffers from known overthinking and repetition issues and is limited to a 32K token output, making the newer GLM-5.3-Flash a more efficient choice for most API work.written, reviewed, cited
  • The 9B GLM-4.6V-Flash model runs locally on a 12 GB gaming GPU or a 24 GB MacBook Pro.
  • GLM-4.6V-FlashX costs $0.04 per million input tokens and $0.4 per million output tokens on the API.
  • Z.ai suggests using top_p 0.6, top_k 2, temperature 0.8, and a repetition penalty of 1.1 to mitigate loops.
GLM-4.6V Arrives: Seamless Integration of Visual Perception and Action Execution | Communeify
communeify.com2026-07-17about the GLM 4.6V family1,568 words
GLM-4.6V launches two versions: a 106B cloud model and a 9B local model, both extending context windows to 128k tokens. The series features native multimodal function calling, enabling direct image-to-tool execution without text conversion. While llama.cpp support for GLM 4.5V/4.6V is still in the draft stage, developers can track progress via GitHub Pull Request #16600.written, reviewed, cited
  • The 9B Flash version achieves an OSWorld score of 21.1, demonstrating strong agentic performance despite its lightweight parameter count.
  • Frontend Replication & Visual Editing allows users to modify generated code using natural language commands like 'make the button bigger'.
  • The 9B Flash model scores 89.2 on the AI2D benchmark, outperforming the larger 106B version which scores 88.8.
GLM-4.6V-Flash Specs & Benchmarks | TheModelverse | The Modelverse
themodelverse.in521 words
GLM-4.6V-Flash features a 128,000 token window and was released on Dec 8, 2025. The model offers open weights with $0 per-token API licensing fees for self-hosting.generated, cited
GLM-4.6V — specs, benchmarks & price | AI/TLDR
ai-tldr.devabout the GLM 4.6V family935 words
GLM-4.6V-Flash vs GLM-4.7-Flash — Pricing, Benchmarks & Performance Compared
anotherwrapper.com
GLM-4.6V: Specs & Benchmarks | LLMLearner
llmlearner.com2025-12-08about the GLM 4.6V family667 words
GLM-4.6V Review: Free 9B Flash Beats Qwen2-VL? [Benchmarks]
binaryverseai.com2025-12-09about the GLM 4.6V family2,737 words
GLM 4.6V Review: Best Open-Source Multimodal Model
blog.siray.ai2025-12-09about the GLM 4.6V family1,507 words

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).

Similar models

3 models

Similar models

  • Qwen3.5-9B · 9.65B
    same size band (L), same task (image-text-to-text), more providers, permissive license
  • Qwen3-VL-8B-Instruct · 8.77B
    same task (image-text-to-text), more providers, permissive license
  • Llama-3.2-11B-Vision-Instruct · 10.7B
    same size band (L), same task (image-text-to-text), fewer providers

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:05 UTC · Last verified 2026-10-25

Follow GLM-4.6V-Flash

Watch this model

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