Context Length Exceeded Error On Fresh Install and First Session
Context Length Exceeded Error
diagnostics_20260413_1.zip
Crash diagnostics
diagnostics_20260413_5.zip
To Reproduce
Steps to reproduce the behavior:
- Go to 'Popular question about government and geese ranking'
- Recieve Context Length Exceeded Error
Expected behavior
A comedic response as to how the governments work and how they would rank as a goose
Screenshots
---
System Info
- OS & Arch: windows 6.2.9200 x86_64 Version. Windows 11 10.0.26200 Build 26200
- Interface: UI
- Version: 1.30.0
- Extensions enabled: analyze, apps, developer, skills, todo, Extension Manager, summon, tom
- Provider & Model: local – unsloth/gemma-4-26B-A4B-it-GGUF:Q4_K_M
- GPU: AMD Radeon RX 7900 XTX 24GB
- CPU: AMD Ryzen 9 7950X 16-Core Processor
- Motherboard: Gigabyte B650 EAGLE AX
- BIOS Version: F39, 2026-02-06
Additional context
It seems to get stuck in the loop of Session Continuation Summary or it cannot generate a response because it does not have tokens, which it should have because it is only the first session.
If I try a new chat with the same question of "Describe in detail how various forms of government works and rank each by units of geese." It repeats the same behavior described.
Trying another Model to see if that solves anything. Switching to Llama-3.2-3B-Instruct-GGUF:Q4_K_M generates a correct answer at least.
Wondering if maybe Gemma-4-26B-A4B-it-GGUF:Q4_K_M overcomplicates the question so much that it cannot comprehend the question, but that's just a wild goose guess.
Trying to switch back to Gemma-4-26B-A4B-it-GGUF:Q4_K_M, maybe it got stuck in a weird loop. It seems to generate an answer. However now I'm curious as to why the first attempt gave me a Context Length Exceeded Error. And I think it crashed (See second image). Trying another attempt just to see what happens. If we crash again we can assume that there is a bug. If we do not generate an answer we can also assume that there is a bug. It crashed again, so there is a crashing bug, when using Gemma-4-26B-A4B-it-GGUF:Q4_K_M. Adding a separate diagnostic for when it crashes.
Context Length Exceeded Error On Fresh Install and First Session
Context Length Exceeded Error
diagnostics_20260413_1.zip
Crash diagnostics
diagnostics_20260413_5.zip
To Reproduce
Steps to reproduce the behavior:
Expected behavior
A comedic response as to how the governments work and how they would rank as a goose
Screenshots
System Info
Additional context
It seems to get stuck in the loop of Session Continuation Summary or it cannot generate a response because it does not have tokens, which it should have because it is only the first session.
If I try a new chat with the same question of "Describe in detail how various forms of government works and rank each by units of geese." It repeats the same behavior described.
Trying another Model to see if that solves anything. Switching to Llama-3.2-3B-Instruct-GGUF:Q4_K_M generates a correct answer at least.
Wondering if maybe Gemma-4-26B-A4B-it-GGUF:Q4_K_M overcomplicates the question so much that it cannot comprehend the question, but that's just a wild goose guess.
Trying to switch back to Gemma-4-26B-A4B-it-GGUF:Q4_K_M, maybe it got stuck in a weird loop. It seems to generate an answer. However now I'm curious as to why the first attempt gave me a Context Length Exceeded Error. And I think it crashed (See second image). Trying another attempt just to see what happens. If we crash again we can assume that there is a bug. If we do not generate an answer we can also assume that there is a bug. It crashed again, so there is a crashing bug, when using Gemma-4-26B-A4B-it-GGUF:Q4_K_M. Adding a separate diagnostic for when it crashes.