model family
GLM-OCR logo

GLM-OCR

Published by zai-org. 1 member measured on real machines, across 3 builds. Every number below was signed by the machine that produced it — none of it is from a model card.

what it takes, what it returns
image + text → text

Declared by the network’s catalog, not measured — a signature says what this family would be asked. Whether it answered is the signed capability flags on each member.

3 of 40 builds measured · 37 still open
what each build proved
base · Q4_K_M
base · AlphaOxO
base · axteam
base · Q4_K_M · Structured output — provedbase · Q4_K_M · Sustained output — asked, failedbase · Q4_K_M · Vision — provedbase · Q4_K_M · Counts in images — asked, failedbase · Q4_K_M · Multiple images at once — provedbase · Q4_K_M · Spatial reasoning — asked, failedbase · Q4_K_M · Writes code — provedbase · Q4_K_M · Reasons about code — provedbase · AlphaOxO · Structured output — provedbase · AlphaOxO · Sustained output — asked, failedbase · AlphaOxO · Vision — provedbase · AlphaOxO · Counts in images — asked, failedbase · AlphaOxO · Multiple images at once — asked, failedbase · AlphaOxO · Spatial reasoning — asked, failedbase · AlphaOxO · Writes code — provedbase · AlphaOxO · Reasons about code — asked, failedbase · axteam · Structured output — provedbase · axteam · Sustained output — provedbase · axteam · Vision — provedbase · axteam · Counts in images — provedbase · axteam · Multiple images at once — provedbase · axteam · Spatial reasoning — provedbase · axteam · Writes code — provedbase · axteam · Reasons about code — proved
Structured output
Sustained output
Vision
Counts in images
Multiple images at once
Spatial reasoning
Writes code
Reasons about code
GLM-OCR BASE3 builds measured →
hf.co/mradermacher/GLM-OCR-GGUF:Q4_K_Mbase · Q4_K_M · mradermacher · ollama · 2 machines · 891M
4Krecalled
—served
106.1best tok/s
106.1 tok/s · unnamed machine13.0 tok/s · x86 · CPU · Hyderabad
13.0 → 106.1 tok/s across 2 machines · ×8.2 spreadunnamed machinex86 · CPU
recalled 4k⌐ declared 128k · unsignedadvertises 32× what it proved
AlphaOxO/GLM-OCR-NVFP4base · AlphaOxO · vllm · 1 machine
4Krecalled
—served
—best tok/s
axteam/GLM-OCR-FP8-dynamicbase · axteam · vllm · 1 machine
4Krecalled
—served
—best tok/s

Builds this network knows of

Declared, not measured — this is what the catalog says exists and what it would cost a machine to hold. Nothing here is signed, and none of it claims the build works.

how much memory does your machine have

40 of 40 builds fit a 16 GB machine — budgeting 70% of it, or 11.2 GiB, because the cache and the runtime want the rest.

  • tiny-random/glm-ocrBF16vllmderived by tiny-random
    0.0 GiB
  • yujiepan/glm-ocr-tiny-randomBF16vllmderived by yujiepan
    0.0 GiB
  • huggingface-mic/GLM-OCR-micFP32vllmderived by huggingface-mic
    0.0 GiB
  • hf.co/mradermacher/GLM-OCR-GGUF:Q2_KQ2_Kollamaquantized by mradermacher
    0.3 GiB
  • hf.co/mradermacher/GLM-OCR-i1-GGUF:Q2_KQ2_Kollamaderived by mradermacher
    0.3 GiB
  • hf.co/mradermacher/GLM-OCR-i1-GGUF:IQ3_XSIQ3_XSollamaderived by mradermacher
    0.4 GiB
  • hf.co/mradermacher/GLM-OCR-GGUF:Q3_K_SQ3_K_Sollamaquantized by mradermacher
    0.4 GiB
  • hf.co/mradermacher/GLM-OCR-i1-GGUF:IQ3_SIQ3_Sollamaderived by mradermacher
    0.4 GiB
  • hf.co/mradermacher/GLM-OCR-GGUF:Q3_K_LQ3_K_Lollamaquantized by mradermacher
    0.4 GiB
  • hf.co/mradermacher/GLM-OCR-i1-GGUF:Q4_0Q4_0ollamaderived by mradermacher
    0.5 GiB
  • hf.co/mradermacher/GLM-OCR-GGUF:Q4_K_SQ4_K_Sollamaquantized by mradermacher
    0.5 GiB
  • hf.co/mradermacher/GLM-OCR-GGUF:Q4_K_MQ4_K_Mollamameasured herequantized by mradermacher
    0.5 GiB
  • hf.co/imbenwolf/GLM-OCR-Q4_K_M-GGUF:Q4_K_MQ4_K_Mollamaderived by imbenwolf
    0.5 GiB
  • hf.co/mradermacher/GLM-OCR-i1-GGUF:Q4_K_MQ4_K_Mollamaderived by mradermacher
    0.5 GiB
  • hf.co/nopesadly/GLM-OCR-Q4_K_M.gguf:Q4_K_MQ4_K_Mollamaderived by nopesadly
    0.5 GiB
  • matthewdubois/GLM-OCR-3bit-mlx3BITmlxquantized by matthewdubois
    0.7 GiB
  • hf.co/cstr/glm-ocr-crispembed-GGUF:Q4_KQ4_Kollamaderived by cstr
    0.8 GiB
  • hf.co/ggml-org/GLM-OCR-GGUF:Q8_0Q8_0ollamaquantized by ggml-org
    0.9 GiB
  • hf.co/42ailab/GLM-OCR-GGUF:Q8_0Q8_0ollamaderived by 42ailab
    0.9 GiB
  • hf.co/fabriziosalmi/GLM-OCR-ita-GGUF:Q8_0Q8_0ollamaderived by fabriziosalmi
    0.9 GiB
  • hf.co/RockMan256/glm-ocr-rc-GGUF:Q8_0Q8_0ollamaderived by RockMan256
    0.9 GiB
  • hf.co/cstr/glm-ocr-crispembed-GGUF:Q8_0Q8_0ollamaderived by cstr
    1.1 GiB
  • fabriziosalmi/GLM-OCR-ita-mlx-8bit8BITmlxderived by fabriziosalmi
    1.1 GiB
  • EZCon/GLM-OCR-4bit-mlx4BITmlxquantized by EZCon
    1.2 GiB
  • mlx-community/GLM-OCR-4bit4BITmlxderived by mlx-community
    1.2 GiB
  • AlphaOxO/GLM-OCR-NVFP4W4A16vllmmeasured herequantized by AlphaOxO
    1.3 GiB
  • EZCon/GLM-OCR-4bit-g32-mxfp4-mixed_4_8-mlx4BITmlxderived by EZCon
    1.3 GiB
  • mlx-community/GLM-OCR-6bit6BITmlxderived by mlx-community
    1.3 GiB
  • EZCon/GLM-OCR-8bit-mlx8BITmlxquantized by EZCon
    1.5 GiB
  • mlx-community/GLM-OCR-8bit8BITmlxderived by mlx-community
    1.5 GiB
  • npario/GLM-OCR-8bit8BITmlxderived by npario
    1.5 GiB
  • RepublicOfKorokke/GLM-OCR-oQ8-fp168BITmlxderived by RepublicOfKorokke
    1.5 GiB
  • vastai-ais/GLM-OCR-FP8FP8vllmderived by vastai-ais
    1.6 GiB
  • hf.co/ggml-org/GLM-OCR-GGUF:F16F16ollamaquantized by ggml-org
    1.7 GiB
  • axteam/GLM-OCR-FP8-dynamicFP8vllmmeasured herequantized by axteam
    2.0 GiB
  • Saidakmal/GLM-OCR-uzBF16vllmderived by Saidakmal
    2.1 GiB
  • sherif1313/Arabic-GLM-OCR-v1FP16vllmderived by sherif1313
    2.1 GiB
  • sherif1313/Arabic-GLM-OCR-v2FP16vllmderived by sherif1313
    2.1 GiB
  • hf.co/cstr/glm-ocr-crispembed-GGUF:F16F16ollamaderived by cstr
    2.4 GiB
  • larshiakzemil/glm-ocr-farsiFP32vllmderived by larshiakzemil
    4.1 GiB

Sizes are what the catalog declares, not what a machine measured — fitting on disk is necessary and not sufficient. Nobody here has run these on your machine, and this narrows the candidates rather than certifying one.

the same builds, by member
glm-ocr-crispembed-GGUF
hf.co/cstr/glm-ocr-crispembed-GGUF:Q4_KQ4_Kollama848 MiB on diskderived by cstr
hf.co/cstr/glm-ocr-crispembed-GGUF:Q8_0Q8_0ollama1.1 GiB on diskderived by cstr
hf.co/cstr/glm-ocr-crispembed-GGUF:F16F16ollama2.4 GiB on diskderived by cstr
GLM-OCR BASE
hf.co/mradermacher/GLM-OCR-GGUF:Q2_KQ2_Kollama344 MiB on diskquantized by mradermacher
hf.co/mradermacher/GLM-OCR-GGUF:Q3_K_SQ3_K_Sollama402 MiB on diskquantized by mradermacher
hf.co/mradermacher/GLM-OCR-GGUF:Q3_K_LQ3_K_Lollama451 MiB on diskquantized by mradermacher
hf.co/mradermacher/GLM-OCR-GGUF:Q4_K_SQ4_K_Sollama506 MiB on diskquantized by mradermacher
hf.co/mradermacher/GLM-OCR-GGUF:Q4_K_MQ4_K_Mollama523 MiB on diskquantized by mradermacher
matthewdubois/GLM-OCR-3bit-mlx3BITmlx728 MiB on diskquantized by matthewdubois
hf.co/ggml-org/GLM-OCR-GGUF:Q8_0Q8_0ollama906 MiB on diskquantized by ggml-org
EZCon/GLM-OCR-4bit-mlx4BITmlx1.2 GiB on diskquantized by EZCon
AlphaOxO/GLM-OCR-NVFP4W4A16vllm1.3 GiB on diskquantized by AlphaOxO
EZCon/GLM-OCR-8bit-mlx8BITmlx1.5 GiB on diskquantized by EZCon
hf.co/ggml-org/GLM-OCR-GGUF:F16F16ollama1.7 GiB on diskquantized by ggml-org
axteam/GLM-OCR-FP8-dynamicFP8vllm2.0 GiB on diskquantized by axteam
GLM-OCR-GGUF
hf.co/42ailab/GLM-OCR-GGUF:Q8_0Q8_0ollama906 MiB on diskderived by 42ailab
GLM-OCR-4bit-g32-mxfp4-mixed_4_8-mlx
EZCon/GLM-OCR-4bit-g32-mxfp4-mixed_4_8-mlx4BITmlx1.3 GiB on diskderived by EZCon
GLM-OCR-ita-GGUF
hf.co/fabriziosalmi/GLM-OCR-ita-GGUF:Q8_0Q8_0ollama906 MiB on diskderived by fabriziosalmi
fabriziosalmi/GLM-OCR-ita-mlx-8bit8BITmlx1.1 GiB on diskderived by fabriziosalmi
GLM-OCR-mic
huggingface-mic/GLM-OCR-micFP32vllm33 MiB on diskderived by huggingface-mic
GLM-OCR-Q4_K_M-GGUF
hf.co/imbenwolf/GLM-OCR-Q4_K_M-GGUF:Q4_K_MQ4_K_Mollama523 MiB on diskderived by imbenwolf
glm-ocr-farsi
larshiakzemil/glm-ocr-farsiFP32vllm4.1 GiB on diskderived by larshiakzemil
GLM-OCR-bf16
mlx-community/GLM-OCR-4bit4BITmlx1.2 GiB on diskderived by mlx-community
mlx-community/GLM-OCR-6bit6BITmlx1.3 GiB on diskderived by mlx-community
mlx-community/GLM-OCR-8bit8BITmlx1.5 GiB on diskderived by mlx-community
GLM-OCR-i1-GGUF
hf.co/mradermacher/GLM-OCR-i1-GGUF:Q2_KQ2_Kollama344 MiB on diskderived by mradermacher
hf.co/mradermacher/GLM-OCR-i1-GGUF:IQ3_XSIQ3_XSollama391 MiB on diskderived by mradermacher
hf.co/mradermacher/GLM-OCR-i1-GGUF:IQ3_SIQ3_Sollama402 MiB on diskderived by mradermacher
hf.co/mradermacher/GLM-OCR-i1-GGUF:Q4_0Q4_0ollama504 MiB on diskderived by mradermacher
hf.co/mradermacher/GLM-OCR-i1-GGUF:Q4_K_MQ4_K_Mollama523 MiB on diskderived by mradermacher
GLM-OCR-Q4_K_M.gguf
hf.co/nopesadly/GLM-OCR-Q4_K_M.gguf:Q4_K_MQ4_K_Mollama523 MiB on diskderived by nopesadly
GLM-OCR-bf16
npario/GLM-OCR-8bit8BITmlx1.5 GiB on diskderived by npario
GLM-OCR-oQ8-fp16
RepublicOfKorokke/GLM-OCR-oQ8-fp168BITmlx1.5 GiB on diskderived by RepublicOfKorokke
GLM-OCR-uz
Saidakmal/GLM-OCR-uzBF16vllm2.1 GiB on diskderived by Saidakmal
Arabic-GLM-OCR-v1
sherif1313/Arabic-GLM-OCR-v1FP16vllm2.1 GiB on diskderived by sherif1313
Arabic-GLM-OCR-v2
sherif1313/Arabic-GLM-OCR-v2FP16vllm2.1 GiB on diskderived by sherif1313
glm-ocr
tiny-random/glm-ocrBF16vllm3 MiB on diskderived by tiny-random
GLM-OCR-FP8
vastai-ais/GLM-OCR-FP8FP8vllm1.6 GiB on diskderived by vastai-ais
glm-ocr-tiny-random
yujiepan/glm-ocr-tiny-randomBF16vllm3 MiB on diskderived by yujiepan
glm-ocr-rc-GGUF
hf.co/RockMan256/glm-ocr-rc-GGUF:Q8_0Q8_0ollama906 MiB on diskderived by RockMan256
ask your own gram

Your concierge can answer this from the network rather than from this page. The ask carries the operator id, so it resolves to a capability and an engine — not to a name somebody has to recognise.

What has this network measured about the GLM-OCR family (epn.family.glm-ocr), and can my gram run any of it? See /inferences/glm-ocr

open a gramx with this ask →epn.family.glm-ocr