github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-17 05:36:16 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #13003] Ollam 0.12.10 Qwen3-VL:235b-instruct-cloud can't handle more than 5000 Words in German #70669

New Issue
Open
opened 2026-05-04 22:28:10 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-05-04 22:28:10 -05:00
Owner
Copy Link

Originally created by @sto1 on GitHub (Nov 7, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13003

What is the issue?

when I send a Context window from my RAG with more than 5000 Words, the Qwen cloud models give no answer! No output on the ollama server, no response, nothing to see. I expect all Cloud models from qwen3 to behave like this. I tested instruct, and the "normal" one.

Relevant log output

Nothing to see just wait. all other models work, except small qwen3 models on my local server, but this bug is already reported.

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.12.10

Originally created by @sto1 on GitHub (Nov 7, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13003 ### What is the issue? when I send a Context window from my RAG with more than 5000 Words, the Qwen cloud models give no answer! No output on the ollama server, no response, nothing to see. I expect all Cloud models from qwen3 to behave like this. I tested instruct, and the "normal" one. ### Relevant log output ```shell Nothing to see just wait. all other models work, except small qwen3 models on my local server, but this bug is already reported. ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.12.10
GiteaMirror added the bug label 2026-05-04 22:28:10 -05:00
GiteaMirror commented 2026-05-04 22:28:16 -05:00
Author
Owner
Copy Link

@sto1 commented on GitHub (Nov 8, 2025):

5'000 German words is about 7'500 token! It works up to 7'500 token. if more token, nothing is coming back. I use other models with 50'000 words and some models gpt-oss:120b-cloud up to 100'000 words.

qwen3-vl:235b-instruct-cloud works fine! Have only issues with qwen3-vl

<!-- gh-comment-id:3506076799 --> @sto1 commented on GitHub (Nov 8, 2025): 5'000 German words is about 7'500 token! It works up to 7'500 token. if more token, nothing is coming back. I use other models with 50'000 words and some models gpt-oss:120b-cloud up to 100'000 words. qwen3-vl:235b-instruct-cloud works fine! Have only issues with qwen3-vl
GiteaMirror commented 2026-05-04 22:28:23 -05:00
Author
Owner
Copy Link

@shtshang commented on GitHub (Nov 8, 2025):

I try to upload two big pictures to Qwen3-VL:235b-instruct-cloud, and it will return nothing.
But one is OK, and uploading several small pictures is also OK.
It seems that the context is a little small.

<!-- gh-comment-id:3506446603 --> @shtshang commented on GitHub (Nov 8, 2025): I try to upload two big pictures to Qwen3-VL:235b-instruct-cloud, and it will return nothing. But one is OK, and uploading several small pictures is also OK. It seems that the context is a little small.
GiteaMirror commented 2026-05-04 22:28:26 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 8, 2025):

https://docs.ollama.com/context-length

<!-- gh-comment-id:3506712419 --> @rick-github commented on GitHub (Nov 8, 2025): https://docs.ollama.com/context-length
GiteaMirror commented 2026-05-04 22:28:28 -05:00
Author
Owner
Copy Link

@sto1 commented on GitHub (Nov 8, 2025):

I use RAG with llama-index and in all versions (ollama, ollama local, OpenAI, gemini) the Context window is maxed. As said, with the same configuration, am able to go much larger (even gemma3), with the token size.

<!-- gh-comment-id:3506737442 --> @sto1 commented on GitHub (Nov 8, 2025): I use RAG with llama-index and in all versions (ollama, ollama local, OpenAI, gemini) the Context window is maxed. As said, with the same configuration, am able to go much larger (even gemma3), with the token size.
GiteaMirror commented 2026-05-04 22:28:29 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 8, 2025):

Server log will help in debugging.

<!-- gh-comment-id:3506742794 --> @rick-github commented on GitHub (Nov 8, 2025): [Server log](https://docs.ollama.com/troubleshooting) will help in debugging.
GiteaMirror commented 2026-05-04 22:28:29 -05:00
Author
Owner
Copy Link

@sto1 commented on GitHub (Nov 8, 2025):

first call with 5000 words worked at 20:10:30
Second call with 10'000 words back no answer
third call with 5000 words worked at 20:14:50

nothing from the second call at all!
I don't know how to log the cloud!
Thanks for your great work!!

Windows PowerShell
Copyright (C) Microsoft Corporation. Alle Rechte vorbehalten.

Installieren Sie die neueste PowerShell für neue Funktionen und Verbesserungen! https://aka.ms/PSWindows

Das Laden von persönlichen und Systemprofilen dauerte 2936 ms.
(base) PS C:\Users\storc> ollama_start.bat
Starte Ollama mit Nvidia GPU (Index 0)
time=2025-11-08T20:04:40.967+01:00 level=INFO source=routes.go:1525 msg="server config" env="map[CUDA_VISIBLE_DEVICES:0 GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:true OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE:f16 OLLAMA_LLM_LIBRARY:cuda_v12 OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:I:\Users\storc\.ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[chrome-extension://* moz-extension://* safari-web-extension://* https://hollama.fernando.is http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]"
time=2025-11-08T20:04:41.135+01:00 level=INFO source=images.go:522 msg="total blobs: 77"
time=2025-11-08T20:04:41.138+01:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0"
time=2025-11-08T20:04:41.140+01:00 level=INFO source=routes.go:1578 msg="Listening on [::]:11434 (version 0.12.10)"
time=2025-11-08T20:04:41.141+01:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2025-11-08T20:04:41.148+01:00 level=INFO source=server.go:400 msg="starting runner" cmd="C:\Users\storc\AppData\Local\Programs\Ollama\ollama.exe runner --ollama-engine --port 56915"
time=2025-11-08T20:04:43.180+01:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-cd38eb1c-290a-15c2-d573-b2c87845fde3 filter_id="" library=CUDA compute=8.6 name=CUDA0 description="NVIDIA GeForce RTX 3060" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:2d:00.0 type=discrete total="12.0 GiB" available="11.8 GiB"
time=2025-11-08T20:04:43.180+01:00 level=INFO source=routes.go:1619 msg="entering low vram mode" "total vram"="12.0 GiB" threshold="20.0 GiB"
[GIN] 2025/11/08 - 20:08:24 | 200 | 12.539ms | 127.0.0.1 | POST "/api/show"
[GIN] 2025/11/08 - 20:08:36 | 200 | 12.7450661s | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:09:57 | 200 | 27.7457957s | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:10:30 | 200 | 32.6797716s | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:12:04 | 200 | 29.934087s | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:12:06 | 200 | 838.4495ms | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:14:50 | 200 | 29.0411745s | 127.0.0.1 | POST "/api/chat"
[GIN] 2025/11/08 - 20:15:30 | 200 | 39.3671621s | 127.0.0.1 | POST "/api/chat"

<!-- gh-comment-id:3506814871 --> @sto1 commented on GitHub (Nov 8, 2025): first call with 5000 words worked at 20:10:30 Second call with 10'000 words back no answer third call with 5000 words worked at 20:14:50 nothing from the second call at all! I don't know how to log the cloud! Thanks for your great work!! Windows PowerShell Copyright (C) Microsoft Corporation. Alle Rechte vorbehalten. Installieren Sie die neueste PowerShell für neue Funktionen und Verbesserungen! https://aka.ms/PSWindows Das Laden von persönlichen und Systemprofilen dauerte 2936 ms. (base) PS C:\Users\storc> ollama_start.bat Starte Ollama mit Nvidia GPU (Index 0) time=2025-11-08T20:04:40.967+01:00 level=INFO source=routes.go:1525 msg="server config" env="map[CUDA_VISIBLE_DEVICES:0 GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:true OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://0.0.0.0:11434 OLLAMA_INTEL_GPU:false OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE:f16 OLLAMA_LLM_LIBRARY:cuda_v12 OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:I:\\Users\\storc\\.ollama OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[chrome-extension://* moz-extension://* safari-web-extension://* https://hollama.fernando.is http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]" time=2025-11-08T20:04:41.135+01:00 level=INFO source=images.go:522 msg="total blobs: 77" time=2025-11-08T20:04:41.138+01:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0" time=2025-11-08T20:04:41.140+01:00 level=INFO source=routes.go:1578 msg="Listening on [::]:11434 (version 0.12.10)" time=2025-11-08T20:04:41.141+01:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2025-11-08T20:04:41.148+01:00 level=INFO source=server.go:400 msg="starting runner" cmd="C:\\Users\\storc\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --ollama-engine --port 56915" time=2025-11-08T20:04:43.180+01:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-cd38eb1c-290a-15c2-d573-b2c87845fde3 filter_id="" library=CUDA compute=8.6 name=CUDA0 description="NVIDIA GeForce RTX 3060" libdirs=ollama,cuda_v12 driver=12.9 pci_id=0000:2d:00.0 type=discrete total="12.0 GiB" available="11.8 GiB" time=2025-11-08T20:04:43.180+01:00 level=INFO source=routes.go:1619 msg="entering low vram mode" "total vram"="12.0 GiB" threshold="20.0 GiB" [GIN] 2025/11/08 - 20:08:24 | 200 | 12.539ms | 127.0.0.1 | POST "/api/show" [GIN] 2025/11/08 - 20:08:36 | 200 | 12.7450661s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:09:57 | 200 | 27.7457957s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:10:30 | 200 | 32.6797716s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:12:04 | 200 | 29.934087s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:12:06 | 200 | 838.4495ms | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:14:50 | 200 | 29.0411745s | 127.0.0.1 | POST "/api/chat" [GIN] 2025/11/08 - 20:15:30 | 200 | 39.3671621s | 127.0.0.1 | POST "/api/chat"
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#70669
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?