Eval bug: MiniMax-H3-GGUF fails to load up

[dogunbound]

Name and Version

[dogunbound@dog-arch bin]$ ./llama-cli --version
version: 0.3.0-dev (build 10675, commit 90c26fc)
built with IntelLLVM 2026.0.0 for Linux x86_64

Operating systems

Linux

GGML backends

SYCL

Hardware

[dogunbound@dog-arch bin]$ sycl-ls
[level_zero:gpu][level_zero:0] Intel(R) oneAPI Unified Runtime over Level-Zero V2, Intel(R) Arc(TM) Pro B70 Graphics 20.2.0 [1.17.39395]
[level_zero:gpu][level_zero:1] Intel(R) oneAPI Unified Runtime over Level-Zero V2, Intel(R) Arc(TM) A750 Graphics 12.55.8 [1.17.39395]
[opencl:cpu][opencl:0] Intel(R) OpenCL, 12th Gen Intel(R) Core(TM) i7-12700K OpenCL 3.0 (Build 0) [2026.21.3.0.31_160000]
[opencl:gpu][opencl:1] Intel(R) OpenCL Graphics, Intel(R) Arc(TM) Pro B70 Graphics OpenCL 3.0 NEO  [26.31.39395]
[opencl:gpu][opencl:2] Intel(R) OpenCL Graphics, Intel(R) Arc(TM) A750 Graphics OpenCL 3.0 NEO  [26.31.39395]

Models

Any of the minimax models here: https://huggingface.co/unsloth/MiniMax-H3-GGUF

Problem description & steps to reproduce

[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF


Loading model... |0.00.327.002 E llama_model_load: error loading model: unknown model architecture: ''
0.00.327.004 E llama_model_load_from_file_impl: failed to load model
0.00.327.027 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.330.133 E llama_model_load: error loading model: unknown model architecture: ''
0.00.330.134 E llama_model_load_from_file_impl: failed to load model
0.00.330.135 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/qwen3vl_32b_minimax_h3-Q4_K_M.gguf'
0.00.330.139 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/qwen3vl_32b_minimax_h3-Q4_K_M.gguf'
0.00.330.950 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF ^C
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF:Q8_0


Loading model... |0.00.343.248 E llama_model_load: error loading model: unknown model architecture: ''
0.00.343.251 E llama_model_load_from_file_impl: failed to load model
0.00.343.281 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.345.658 E llama_model_load: error loading model: unknown model architecture: ''
0.00.345.660 E llama_model_load_from_file_impl: failed to load model
0.00.345.661 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.345.666 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.346.387 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF:^C_0
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_XL -mg 0 -sm none -c 512000^C
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF:pruned-Q8_0


Loading model... |0.00.335.896 E llama_model_load: error loading model: unknown model architecture: ''
0.00.335.898 E llama_model_load_from_file_impl: failed to load model
0.00.335.920 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.339.836 E llama_model_load: error loading model: unknown model architecture: ''
0.00.339.844 E llama_model_load_from_file_impl: failed to load model
0.00.339.848 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.339.862 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.341.807 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF:pruned-Q8_0^C
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_
minimax_h3_fl2va_pruned-Q2_K.gguf        minimax_h3_fl2va_pruned-Q8_0.gguf        minimax_h3_ref2va_pruned-Q4_K.gguf
minimax_h3_fl2va_pruned-Q3_K.gguf        minimax_h3_fl2va_pruned-UD-Q2_K_XL.gguf  minimax_h3_ref2va_pruned-Q5_0.gguf
minimax_h3_fl2va_pruned-Q4_K.gguf        minimax_h3_fl2va_pruned-UD-Q3_K_XL.gguf  minimax_h3_ref2va_pruned-Q6_K.gguf
minimax_h3_fl2va_pruned-Q5_0.gguf        minimax_h3_ref2va_pruned-Q2_K.gguf       minimax_h3_ref2va_pruned-Q8_0.gguf
minimax_h3_fl2va_pruned-Q6_K.gguf        minimax_h3_ref2va_pruned-Q3_K.gguf       
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf 


Loading model... |0.00.145.444 E llama_model_load: error loading model: unknown model architecture: ''
0.00.145.448 E llama_model_load_from_file_impl: failed to load model
0.00.145.488 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.148.150 E llama_model_load: error loading model: unknown model architecture: ''
0.00.148.154 E llama_model_load_from_file_impl: failed to load model
0.00.148.156 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf'
0.00.148.160 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf'
0.00.148.906 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf  -mg 0 -sm none -c 512000


Loading model... |0.00.193.115 E llama_model_load: error loading model: unknown model architecture: ''
0.00.193.124 E llama_model_load_from_file_impl: failed to load model
0.00.193.209 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.198.498 E llama_model_load: error loading model: unknown model architecture: ''
0.00.198.505 E llama_model_load_from_file_impl: failed to load model
0.00.198.507 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf'
0.00.198.515 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_ref2va_pruned-Q8_0.gguf'
0.00.199.617 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/
assets/                                  minimax_h3_fl2va_pruned-Q8_0.gguf        minimax_h3_ref2va_pruned-Q8_0.gguf
.gitattributes                           minimax_h3_fl2va_pruned-UD-Q2_K_XL.gguf  NOTICE
LICENSE                                  minimax_h3_fl2va_pruned-UD-Q3_K_XL.gguf  qwen3vl_32b_minimax_h3-Q2_K_M.gguf
minimax_h3_fl2va_pruned-Q2_K.gguf        minimax_h3_ref2va_pruned-Q2_K.gguf       qwen3vl_32b_minimax_h3-Q4_K_M.gguf
minimax_h3_fl2va_pruned-Q3_K.gguf        minimax_h3_ref2va_pruned-Q3_K.gguf       README.md
minimax_h3_fl2va_pruned-Q4_K.gguf        minimax_h3_ref2va_pruned-Q4_K.gguf       vae/
minimax_h3_fl2va_pruned-Q5_0.gguf        minimax_h3_ref2va_pruned-Q5_0.gguf       
minimax_h3_fl2va_pruned-Q6_K.gguf        minimax_h3_ref2va_pruned-Q6_K.gguf       
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_
minimax_h3_fl2va_pruned-Q2_K.gguf        minimax_h3_fl2va_pruned-Q8_0.gguf        minimax_h3_ref2va_pruned-Q4_K.gguf
minimax_h3_fl2va_pruned-Q3_K.gguf        minimax_h3_fl2va_pruned-UD-Q2_K_XL.gguf  minimax_h3_ref2va_pruned-Q5_0.gguf
minimax_h3_fl2va_pruned-Q4_K.gguf        minimax_h3_fl2va_pruned-UD-Q3_K_XL.gguf  minimax_h3_ref2va_pruned-Q6_K.gguf
minimax_h3_fl2va_pruned-Q5_0.gguf        minimax_h3_ref2va_pruned-Q2_K.gguf       minimax_h3_ref2va_pruned-Q8_0.gguf
minimax_h3_fl2va_pruned-Q6_K.gguf        minimax_h3_ref2va_pruned-Q3_K.gguf       
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf 


Loading model... |0.00.190.175 E llama_model_load: error loading model: unknown model architecture: ''
0.00.190.179 E llama_model_load_from_file_impl: failed to load model
0.00.190.222 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.193.119 E llama_model_load: error loading model: unknown model architecture: ''
0.00.193.121 E llama_model_load_from_file_impl: failed to load model
0.00.193.122 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.193.127 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.193.759 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf -mg 0 -sm none


Loading model... |0.00.171.799 E llama_model_load: error loading model: unknown model architecture: ''
0.00.171.802 E llama_model_load_from_file_impl: failed to load model
0.00.171.830 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.174.290 E llama_model_load: error loading model: unknown model architecture: ''
0.00.174.292 E llama_model_load_from_file_impl: failed to load model
0.00.174.294 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.174.299 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.174.927 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf -mg 0 -sm none -c 520000


Loading model... |0.00.147.118 E llama_model_load: error loading model: unknown model architecture: ''
0.00.147.127 E llama_model_load_from_file_impl: failed to load model
0.00.147.187 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.151.204 E llama_model_load: error loading model: unknown model architecture: ''
0.00.151.211 E llama_model_load_from_file_impl: failed to load model
0.00.151.213 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.151.220 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.152.071 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -m ~/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf -mg 0 -sm none -c 512000


Loading model... |0.00.179.676 E llama_model_load: error loading model: unknown model architecture: ''
0.00.179.679 E llama_model_load_from_file_impl: failed to load model
0.00.179.700 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.181.166 E llama_model_load: error loading model: unknown model architecture: ''
0.00.181.167 E llama_model_load_from_file_impl: failed to load model
0.00.181.168 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.181.171 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/minimax_h3_fl2va_pruned-Q8_0.gguf'
0.00.181.544 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready
[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3 -mg 0 -sm none -c 512000
0.00.421.226 E common_download_get_hf_plan: no GGUF files found in repository unsloth/MiniMax-H3
error: --model is required

[dogunbound@dog-arch bin]$ ./llama-cli -hf unsloth/MiniMax-H3-GGUF -mg 0 -sm none -c 512000


Loading model... |0.00.345.902 E llama_model_load: error loading model: unknown model architecture: ''
0.00.345.905 E llama_model_load_from_file_impl: failed to load model
0.00.345.927 E common_fit_params: encountered an error while trying to fit params to free device memory: failed to load model
0.00.349.072 E llama_model_load: error loading model: unknown model architecture: ''
0.00.349.073 E llama_model_load_from_file_impl: failed to load model
0.00.349.075 E cmn  common_init_: failed to load model '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/qwen3vl_32b_minimax_h3-Q4_K_M.gguf'
0.00.349.078 E srv    load_model: failed to load model, '/home/dogunbound/.cache/huggingface/hub/models--unsloth--MiniMax-H3-GGUF/snapshots/d629413c2e5b51b38c453668b75ca3b06ca92703/qwen3vl_32b_minimax_h3-Q4_K_M.gguf'
0.00.349.765 E srv  llama_server: exiting due to model loading error
llama_server exited with code 1
Error: the server exited before becoming ready

First Bad Commit

Haven't tried it before.

Relevant log output

Already posted logs in repro steps.

[arthw]

@dogunbound
According to the log info, the root cause is this LLM is not supported by llama.cpp.
You need to report as enhance to llama.cpp to support this LLM.

It's not SYCL backend issue.

Thank you!

[dogunbound]

Eh. If it's not a bug, then no big deal. Thought I may have stumbled upon something. I'll just use a different model.