Failed to load the model such as gemma4-26b-a4b-it , ternary-bonsai-27b , nanbeige-4.2-3b

[dongbutan]

Which version of LM Studio?
Example: LM Studio 0.4.20 (Build 1)

Which operating system?
Windows 11

What is the bug?
A clear and concise description of what the bug is.

🥲 Failed to load the model
Engine protocol runtime llama-server for mmVGcSTlX6QYmWmVvCHJxgvD exited before becoming healthy. exitCode=1, signal=null

🥲 Failed to load the model
Engine protocol runtime llama-server for azcrwoPSWiHf2fUo6Ee5dGxA exited before becoming healthy. exitCode=1, signal=null

🥲 Failed to load the model
Engine protocol runtime llama-server for FIsmjZhi7TVojf6FQidUtbrq exited before becoming healthy. exitCode=1, signal=null

CPU:AMD Ryzen 7 8745HS, Radeon 780M, 64GB-DDR5-5400

Screenshots
If applicable, add screenshots to help explain your problem.

Logs

[2026-07-23 10:21:52.879] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:21:52.880] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:21:52.880] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:21:52.881] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:21:52.881] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:21:53.243] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:21:53.244] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:21:53.245] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:21:53.245] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:21:53.246] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:21:53.606] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:21:53.606] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:21:53.607] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:21:53.608] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:21:53.608] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:21:54.434] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=multiplexedLoadModel] Error in channel handler: Error: Engine protocol runtime llama-server for U2QRKr7c4lfsVMUAVV2uYk0w exited before becoming healthy. exitCode=1, signal=null
at ChildProcess._0x44552a (D:\Program Files\LM Studio\resources\app.webpack\main\index.js:243:28)
at Object.onceWrapper (node:events:634:26)
at ChildProcess.emit (node:events:531:35)
at ChildProcess._handle.onexit (node:internal/child_process:293:12)
[2026-07-23 10:21:54.435] [error] Model instance not found: U2QRKr7c4lfsVMUAVV2uYk0w
[2026-07-23 10:25:10.486] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:25:10.487] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:25:10.488] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:25:10.488] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:25:10.489] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:25:10.850] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:25:10.850] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:25:10.851] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:25:10.852] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:25:10.852] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:25:11.211] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:25:11.212] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:25:11.212] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 14.87 GB
Context: 936.03 MB
Total: 15.81 GB
[2026-07-23 10:25:11.213] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:25:11.213] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:25:12.037] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=multiplexedLoadModel] Error in channel handler: Error: Engine protocol runtime llama-server for mmVGcSTlX6QYmWmVvCHJxgvD exited before becoming healthy. exitCode=1, signal=null
at ChildProcess._0x44552a (D:\Program Files\LM Studio\resources\app.webpack\main\index.js:243:28)
at Object.onceWrapper (node:events:634:26)
at ChildProcess.emit (node:events:531:35)
at ChildProcess._handle.onexit (node:internal/child_process:293:12)
[2026-07-23 10:25:12.038] [error] Model instance not found: mmVGcSTlX6QYmWmVvCHJxgvD
[2026-07-23 10:26:14.100] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:14.101] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:14.102] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 7.38 GB
Context: 1.10 GB
Total: 8.48 GB
[2026-07-23 10:26:14.102] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:14.103] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:14.488] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:14.488] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:14.489] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 7.38 GB
Context: 1.10 GB
Total: 8.48 GB
[2026-07-23 10:26:14.490] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:14.490] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:14.850] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:14.851] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:14.851] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 7.38 GB
Context: 1.10 GB
Total: 8.48 GB
[2026-07-23 10:26:14.852] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:14.852] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:15.657] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=multiplexedLoadModel] Error in channel handler: Error: Engine protocol runtime llama-server for azcrwoPSWiHf2fUo6Ee5dGxA exited before becoming healthy. exitCode=1, signal=null
at ChildProcess._0x44552a (D:\Program Files\LM Studio\resources\app.webpack\main\index.js:243:28)
at Object.onceWrapper (node:events:634:26)
at ChildProcess.emit (node:events:531:35)
at ChildProcess._handle.onexit (node:internal/child_process:293:12)
[2026-07-23 10:26:15.658] [error] Model instance not found: azcrwoPSWiHf2fUo6Ee5dGxA
[2026-07-23 10:26:35.977] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:35.977] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:35.979] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 4.57 GB
Context: 287.96 MB
Total: 4.86 GB
[2026-07-23 10:26:35.979] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:35.980] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:36.348] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:36.349] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:36.350] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 4.57 GB
Context: 287.96 MB
Total: 4.86 GB
[2026-07-23 10:26:36.350] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:36.351] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:36.735] [error] [LM Studio] GPU Configuration:
Strategy: evenly
Priority: []
Disabled GPUs: []
Limit weight offload to dedicated GPU Memory: OFF
Offload KV Cache to GPU: ON
[2026-07-23 10:26:36.736] [error] [LM Studio] Live GPU memory info:
No live GPU info available
[2026-07-23 10:26:36.737] [error] [LM Studio] Model load size estimate with raw num offload layers 'max' and context length '8192':
Model: 4.57 GB
Context: 287.96 MB
Total: 4.86 GB
[2026-07-23 10:26:36.737] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-23 10:26:36.737] [error] [LM Studio] Resolved GPU config options:
Num Offload Layers: max
Num CPU Expert Layers: 0
Main GPU: 0
Tensor Split: [0]
Disabled GPUs: []
[2026-07-23 10:26:37.544] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=multiplexedLoadModel] Error in channel handler: Error: Engine protocol runtime llama-server for FIsmjZhi7TVojf6FQidUtbrq exited before becoming healthy. exitCode=1, signal=null
at ChildProcess._0x44552a (D:\Program Files\LM Studio\resources\app.webpack\main\index.js:243:28)
at Object.onceWrapper (node:events:634:26)
at ChildProcess.emit (node:events:531:35)
at ChildProcess._handle.onexit (node:internal/child_process:293:12)
[2026-07-23 10:26:37.545] [error] Model instance not found: FIsmjZhi7TVojf6FQidUtbrq

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

[sabun123]

Running into the same error on Windows 11 with an RTX 3090 24GB, no matter the context size set (small or max).
This only seems to occur with the 27B Q2 version, if running the Ternary Bonsai 27B Q2 G64 this works but is unusably slow on LM Studio (despite GPU offload setup, CUDA installed, and runtime is CUDA 12 llama.cpp (Windows) v2.27.0)

Tested running Bonsai 27B Q2 directly via llama-server.exe in Powershell from the official Prism ML github repo and the GGUF model directly from hugging face, and it works fine and speedily. Issue is in LM studio setup.

System specs:
OS: Windows 11 64bit
GPU: RTX 3090 24GB
CPU: AMD Ryzen 9 3950x 16 cores
RAM: 80GB DDR3 2133Mhz

[sabun123]

Just tried on another machine, Apple M4 Max with 128GB RAM. Model Ternary Bonsai 27B Q2.

Same error:

Failed to load the model
Engine protocol runtime llama-server for dhEbaD+GPQuMOIvkzHL5cMXb exited before becoming healthy. exitCode=1, signal=null"

Latest LM Studio. Version 0.4.20+1 (0.4.20+1)
Latest Mac OS. Tahoe Version 26.5.2 (25F84)

GGUF Runtime: Metal llama.cpp (v2.27.0)

Snippet of the failure from the logs:

  Total: 22.28 GB
[2026-07-24 22:58:38.220] [error] [LM Studio] Not using full context length for VRAM overflow calculations due to single GPU setup. Instead, using '8192' as context length for the calculation. Original context length: '262144'.
[2026-07-24 22:58:38.220] [error] [LM Studio] Strict GPU VRAM cap is OFF: GPU offload layers will not be checked for adjustment
[2026-07-24 22:58:38.221] [error] [LM Studio] Resolved GPU config options:
  Num Offload Layers: max
  Num CPU Expert Layers: 0
  Main GPU: 0
  Tensor Split: [0]
  Disabled GPUs: []
[2026-07-24 22:58:38.345] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=multiplexedLoadModel] Error in channel handler: Error: Engine protocol runtime llama-server for dhEbaD+GPQuMOIvkzHL5cMXb exited before becoming healthy. exitCode=1, signal=null
    at ChildProcess._0x59f59c (/Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:243:28)
    at Object.onceWrapper (node:events:634:26)
    at ChildProcess.emit (node:events:531:35)
    at ChildProcess._handle.onexit (node:internal/child_process:293:12)
[2026-07-24 22:58:38.345] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=countTokens] Error in RPC handler: Error: Model is unloaded.
    at /Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:261:4403
    at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
[2026-07-24 22:58:38.345] [error] [LMSInternal][LMSAuthenticator][Client=LM Studio][Endpoint=countTokens] Error in RPC handler: Error: Model is unloaded.
    at /Applications/LM Studio.app/Contents/Resources/app/.webpack/main/index.js:261:4403
    at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
[2026-07-24 22:58:38.345] [error] Model instance not found: dhEbaD+GPQuMOIvkzHL5cMXb

[dongbutan]

目前仍没有解决。期待回复。

[ArnaudContet]

Ternary Bonsai has been reported successfully tested in b10216 of llama.cpp, which we are now beyond in LM Studio (current runtime v2.28.2 is b10252).