bug: Unreadable Answer, independing on used model

[webnobis]

Version: 0.7.2

Describe the Bug

After success installation on my hardware (see hardware.png), I could choose serveral models, but each model (see for instance mistral on jan_0_7_2.png) gets an unreadable answer (language mixes). What can I do, to could work with the chat? Independing on my OS language (German or English), the chat answers are unreadable.

Steps to Reproduce

1. Installation for Linux on similar hardware (see hardware.png)
2. Install a running model (not to big or the same than in screenshot, see jan_0_7_2.png)
3. Start chat with that model, maybe with "Hi"

Screenshots / Logs

[2025-11-06][20:47:41][webview:info@asset://localhost/%2Fhome%2FUser%2F.local%2Fshare%2FJan%2Fdata%2Fextensions%2F%40janhq%2Fllamacpp-extension%2Fdist%2Findex.js:1908:7][INFO] Calling Tauri command getDevices with arg --list-devices [2025-11-06][20:47:41][tauri_plugin_llamacpp::device][INFO] Getting devices from server at path: "/home/User/.local/share/Jan/data/llamacpp/backends/b6929/linux-vulkan-x64/build/bin/llama-server" [2025-11-06][20:47:41][tauri_plugin_llamacpp::device][INFO] Device list output: Available devices: Vulkan0: Intel(R) Graphics (RPL-P) (7841 MiB, 5992 MiB free)

[2025-11-06][20:47:41][tauri_plugin_llamacpp::device][INFO] Parsing device line: 'Vulkan0: Intel(R) Graphics (RPL-P) (7841 MiB, 5992 MiB free)' [2025-11-06][20:47:41][tauri_plugin_llamacpp::device][INFO] Parsed device - ID: 'Vulkan0', Name: 'Intel(R) Graphics (RPL-P)', Mem: 7841, Free: 5992 [2025-11-06][20:55:32][tauri::manager][DEBUG] Asset logs not found; fallback to logs.html [2025-11-06][20:55:32][tauri::manager][DEBUG] Asset logs not found; fallback to logs/index.html [2025-11-06][20:55:32][tauri::manager][DEBUG] Asset logs not found; fallback to index.html [2025-11-06][20:55:33][app_lib::core::extensions::commands][INFO] get jan extensions, path: "/home/User/.local/share/Jan/data/extensions/extensions.json" [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] modelSize: 3522941088 [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Using ctx_size: 8192 [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::utils][INFO] Received ctx_size parameter: Some(8192) [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::utils][INFO] Received model metadata: {"quantize.imatrix.file": "./imatrix.dat", "quantize.imatrix.chunks_count": "74", "tokenizer.ggml.token_type": "<Array of type Int32 with 32768 elements, data skipped>", "tokenizer.ggml.pre": "default", "general.architecture": "llama", "tokenizer.chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}", "llama.attention.layer_norm_rms_epsilon": "0.00001", "quantize.imatrix.dataset": "group_40.txt", "general.name": "models--mistralai--Mistral-7B-Instruct-v0.3", "llama.rope.dimension_count": "128", "tokenizer.ggml.scores": "<Array of type Float32 with 32768 elements, data skipped>", "llama.context_length": "32768", "tokenizer.ggml.eos_token_id": "2", "llama.block_count": "32", "llama.attention.head_count_kv": "8", "tokenizer.ggml.tokens": "<Array of type String with 32768 elements, data skipped>", "llama.embedding_length": "4096", "quantize.imatrix.entries_count": "224", "llama.vocab_size": "32768", "general.file_type": "12", "tokenizer.ggml.bos_token_id": "1", "llama.feed_forward_length": "14336", "llama.attention.head_count": "32", "llama.rope.freq_base": "1000000", "tokenizer.ggml.unknown_token_id": "0", "tokenizer.ggml.add_eos_token": "false", "tokenizer.ggml.add_bos_token": "true", "general.quantization_version": "2", "tokenizer.ggml.model": "llama"} [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::utils][INFO] Calculated key_len and val_len from embedding_length: 4096 / 32 heads = 128 per head [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::utils][INFO] KV estimate (no SWA detected) -> full: 1073741824 bytes (~1024.00 MB) [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] isModelSupported: Total memory requirement: 4596682912 for /home/User/.local/share/Jan/data/llamacpp/models/Mistral-7B-Instruct-v0_3_Q3_K_M/model.gguf; Got kvCacheSize: 1073741824 from BE [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM reported/calculated (in bytes): 8221884416 [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] System RAM: 16443768832 bytes [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM: 8221884416 bytes [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Usable total memory: 20088672870 bytes [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Usable VRAM: 5933394227 bytes [2025-11-06][20:56:32][tauri_plugin_llamacpp::gguf::commands][INFO] Required: 4596682912 bytes [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] modelSize: 3522941088 [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Using ctx_size: 8192 [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::utils][INFO] Received ctx_size parameter: Some(8192) [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::utils][INFO] Received model metadata: {"tokenizer.ggml.model": "llama", "tokenizer.ggml.scores": "<Array of type Float32 with 32768 elements, data skipped>", "tokenizer.ggml.add_bos_token": "true", "general.quantization_version": "2", "llama.rope.freq_base": "1000000", "tokenizer.ggml.unknown_token_id": "0", "llama.rope.dimension_count": "128", "general.architecture": "llama", "llama.attention.head_count_kv": "8", "general.file_type": "12", "tokenizer.ggml.bos_token_id": "1", "tokenizer.ggml.eos_token_id": "2", "quantize.imatrix.dataset": "group_40.txt", "tokenizer.ggml.add_eos_token": "false", "llama.context_length": "32768", "quantize.imatrix.file": "./imatrix.dat", "general.name": "models--mistralai--Mistral-7B-Instruct-v0.3", "tokenizer.chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}", "tokenizer.ggml.token_type": "<Array of type Int32 with 32768 elements, data skipped>", "llama.vocab_size": "32768", "llama.feed_forward_length": "14336", "quantize.imatrix.entries_count": "224", "llama.attention.head_count": "32", "llama.block_count": "32", "llama.attention.layer_norm_rms_epsilon": "0.00001", "tokenizer.ggml.pre": "default", "tokenizer.ggml.tokens": "<Array of type String with 32768 elements, data skipped>", "llama.embedding_length": "4096", "quantize.imatrix.chunks_count": "74"} [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::utils][INFO] Calculated key_len and val_len from embedding_length: 4096 / 32 heads = 128 per head [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::utils][INFO] KV estimate (no SWA detected) -> full: 1073741824 bytes (~1024.00 MB) [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] isModelSupported: Total memory requirement: 4596682912 for /home/User/.local/share/Jan/data/llamacpp/models/Mistral-7B-Instruct-v0_3_Q3_K_M/model.gguf; Got kvCacheSize: 1073741824 from BE [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM reported/calculated (in bytes): 8221884416 [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] System RAM: 16443768832 bytes [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM: 8221884416 bytes [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Usable total memory: 20088672870 bytes [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Usable VRAM: 5933394227 bytes [2025-11-06][20:56:37][tauri_plugin_llamacpp::gguf::commands][INFO] Required: 4596682912 bytes [2025-11-06][20:56:37][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: GET /health 127.0.0.1 200 [2025-11-06][20:56:37][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv params_from_: Chat format: Content-only [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot get_availabl: id 2 | task -1 | selected slot by LRU, t_last = -1 [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot launch_slot_: id 2 | task 2527 | processing task [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot update_slots: id 2 | task 2527 | new prompt, n_ctx_slot = 8192, n_keep = 0, task.n_tokens = 235 [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot update_slots: id 2 | task 2527 | n_tokens = 0, memory_seq_rm [0, end) [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot update_slots: id 2 | task 2527 | prompt processing progress, n_tokens = 235, batch.n_tokens = 236, progress = 1.000000 [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] slot update_slots: id 2 | task 2527 | prompt done, n_tokens = 235, batch.n_tokens = 236 [2025-11-06][20:56:37][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: GET /health 127.0.0.1 200 [2025-11-06][20:56:37][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /apply-template 127.0.0.1 200 [2025-11-06][20:56:37][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:56:37][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /tokenize 127.0.0.1 200 [2025-11-06][20:56:56][tauri::manager][DEBUG] Asset logs not found; fallback to logs.html [2025-11-06][20:56:56][tauri::manager][DEBUG] Asset logs not found; fallback to logs/index.html [2025-11-06][20:56:56][tauri::manager][DEBUG] Asset logs not found; fallback to index.html [2025-11-06][20:56:57][app_lib::core::extensions::commands][INFO] get jan extensions, path: "/home/User/.local/share/Jan/data/extensions/extensions.json" [2025-11-06][20:57:09][tauri::manager][DEBUG] Asset logs not found; fallback to logs.html [2025-11-06][20:57:09][tauri::manager][DEBUG] Asset logs not found; fallback to logs/index.html [2025-11-06][20:57:09][tauri::manager][DEBUG] Asset logs not found; fallback to index.html [2025-11-06][20:57:10][app_lib::core::extensions::commands][INFO] get jan extensions, path: "/home/User/.local/share/Jan/data/extensions/extensions.json" [2025-11-06][20:57:23][tauri::manager][DEBUG] Asset logs not found; fallback to logs.html [2025-11-06][20:57:23][tauri::manager][DEBUG] Asset logs not found; fallback to logs/index.html [2025-11-06][20:57:23][tauri::manager][DEBUG] Asset logs not found; fallback to index.html [2025-11-06][20:57:24][app_lib::core::extensions::commands][INFO] get jan extensions, path: "/home/User/.local/share/Jan/data/extensions/extensions.json" [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] modelSize: 3522941088 [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Using ctx_size: 8192 [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::utils][INFO] Received ctx_size parameter: Some(8192) [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::utils][INFO] Received model metadata: {"general.architecture": "llama", "tokenizer.ggml.add_bos_token": "true", "general.name": "models--mistralai--Mistral-7B-Instruct-v0.3", "llama.embedding_length": "4096", "tokenizer.ggml.model": "llama", "llama.feed_forward_length": "14336", "general.quantization_version": "2", "tokenizer.ggml.eos_token_id": "2", "tokenizer.ggml.pre": "default", "llama.rope.dimension_count": "128", "quantize.imatrix.entries_count": "224", "quantize.imatrix.dataset": "group_40.txt", "tokenizer.ggml.scores": "<Array of type Float32 with 32768 elements, data skipped>", "quantize.imatrix.chunks_count": "74", "tokenizer.ggml.tokens": "<Array of type String with 32768 elements, data skipped>", "llama.attention.head_count": "32", "llama.vocab_size": "32768", "llama.block_count": "32", "llama.context_length": "32768", "tokenizer.ggml.add_eos_token": "false", "general.file_type": "12", "tokenizer.ggml.bos_token_id": "1", "tokenizer.ggml.unknown_token_id": "0", "llama.attention.layer_norm_rms_epsilon": "0.00001", "tokenizer.ggml.token_type": "<Array of type Int32 with 32768 elements, data skipped>", "llama.attention.head_count_kv": "8", "quantize.imatrix.file": "./imatrix.dat", "tokenizer.chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}", "llama.rope.freq_base": "1000000"} [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::utils][INFO] Calculated key_len and val_len from embedding_length: 4096 / 32 heads = 128 per head [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::utils][INFO] KV estimate (no SWA detected) -> full: 1073741824 bytes (~1024.00 MB) [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] isModelSupported: Total memory requirement: 4596682912 for /home/User/.local/share/Jan/data/llamacpp/models/Mistral-7B-Instruct-v0_3_Q3_K_M/model.gguf; Got kvCacheSize: 1073741824 from BE [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM reported/calculated (in bytes): 8221884416 [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] System RAM: 16443768832 bytes [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Total VRAM: 8221884416 bytes [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Usable total memory: 20088672870 bytes [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Usable VRAM: 5933394227 bytes [2025-11-06][20:57:47][tauri_plugin_llamacpp::gguf::commands][INFO] Required: 4596682912 bytes [2025-11-06][20:57:48][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:57:48][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: GET /health 127.0.0.1 200 [2025-11-06][20:57:48][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:57:48][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /apply-template 127.0.0.1 200 [2025-11-06][20:57:48][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:57:48][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /tokenize 127.0.0.1 200 [2025-11-06][20:58:11][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:11][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: GET /health 127.0.0.1 200 [2025-11-06][20:58:11][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:11][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /apply-template 127.0.0.1 200 [2025-11-06][20:58:11][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:11][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /tokenize 127.0.0.1 200 [2025-11-06][20:58:14][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:14][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: GET /health 127.0.0.1 200 [2025-11-06][20:58:14][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:14][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /apply-template 127.0.0.1 200 [2025-11-06][20:58:14][reqwest::connect][DEBUG] starting new connection: http://localhost:3365/ [2025-11-06][20:58:14][tauri_plugin_llamacpp::commands][INFO] [llamacpp] srv log_server_r: request: POST /tokenize 127.0.0.1 200

Operating System

• [ ] MacOS
• [ ] Windows
• [X] Linux

[github-actions[bot]]

Found 3 possible duplicate issues:

1. https://github.com/janhq/jan/issues/3030
2. https://github.com/janhq/jan/issues/3137
3. https://github.com/janhq/jan/issues/1318

[webnobis]

After reinstallation, the same issue. Each interaction results in unreadable content.

Is there a way to get english or german answers? Maybe do I need only a config change?