Model behaviour issues on Linux — page 2
Issues the maintainers labelled both platform:linux and area:model.
1,695 issues · 522 open · 1,173 resolved (69%) · first seen May 14, 2025
Is this getting better or worse?
This class of problem is still growing. 1383 new reports in the last 90 days vs 176 in the 90 before — +686%. The open backlog peaked at 908 in 2026-07 and sits at 522 today.
Derived by counting each matching issue from its creation date until it closed. First-party
computation over the anthropics/claude-code tracker — the trend is not published
anywhere upstream.
Which builds report this
637 of these reports name the Claude Code build they were running, spanning 145 releases. Heaviest reporters:
- Claude Code v2.1.201
- Claude Code v2.1.170
- Claude Code v2.1.198
- Claude Code v2.1.207
- Claude Code v2.1.241
- Claude Code v2.1.211
- Claude Code v2.1.206
- Claude Code v2.1.233
Report volume, not a defect count: an issue can mention a build without that build having caused the failure. Versions come from the version string in the bug-report template, matched against real published tags.
Do these get fixed?
Median time to close is 46 days across 1,173 closures. Of the 1,173 closures with a recorded reason, 8% were closed as completed and 1,084 as not-planned or duplicate. The most recent completed fix landed in 2026-08.
Workarounds reported by the community
54 of these issues post a workaround someone says worked. The highest-engagement ones:
- [BUG] Can't see Claude's thinking anymore
- …problem as you. Looking through other issues, I found this workaround: Hopefully they fix it in general soon. In many cases, Claude is unusable when the thinking is not displayed. It runs for minutes, burns tokens, and…
Found in the thread of #24285 · still open - [BUG] max_tokens must be greater than thinking.budget_tokens
- workaround is possible by editing settings.json (in my case in ~/.claude/settings.json), by setting the "alwaysThinkingEnabled" to false.
Found in the thread of #8756 · still open - Tool calls intermittently emitted as literal text (stray 'court' + raw <invoke>) instead of executing — mostly Edit/Read (Opus, CLI)
- …k is now access-restricted for us — leaving no model-switch workaround — please consider raising severity above cosmetic. Happy to share a sanitized JSONL excerpt of the dropped-call events. Detector we used (reproduces…
Found in the thread of #64108 · still open - [Bug] Anthropic API Error: Usage credits required for 1M context window with opus-plan model
- …sn't work, and just stops working at 1M Context. A feasible workaround would be to periodically use the /compact command manually.
Found in the thread of #61869 · resolved - [BUG] Settings effortLevel "max" is silently downgraded when user interacts with effort selection UI
- …ot above. Using only one of the variables doesn't fix that. Workaround: - then it shows the effort in the welcome text, in the footer, and as the selected effort in .
Found in the thread of #30726 · still open
Excerpts are quoted from the linked issue so you can judge relevance before opening it. Threads that only ask for a workaround, or report that there is none, are excluded. These are community suggestions, not official guidance from Anthropic — read the full thread before applying one.
Matching issues 101–200
Ranked by community engagement (reactions weighted over comments). Listing the top 300 of 1,695 matches; the statistics above cover all 1,695.
- [BUG] API error on Opus 4.1 and Opus 4 models
- [BUG] "parent" are double-stringified in Notion-MCP tool calls → server throws “Expected object, received string”
- [BUG] Claude 4 (or since its release) is BURNING USAGE LIMITS x10 TIMES FASTER
- [P0 CRITICAL] Repeated false-positive “cybersecurity topic” blocks on legitimate commercial ops work involving Grokbot / xAI (Opus 4.7)
- [BUG] Sonnet 4.6 1m missing from Max plan
- [BUG] Native macOS/Linux builds (>=v2.1.117) removed Glob/Grep tools, but model guidance still points to them — Claude reaches for tools that no longer exist
- [Feature Request] Support for Fable5 model
- [Bug] Usage Policy false positive blocks legitimate defensive security research
- [BUG] Every single attempt at reviewing my codebase triggers fall back to opus 4.8
- False positive: Usage Policy block on lattice gauge theory question Fable 5, v2.1.170) [BUG] Request ID: req_011Cbt6Zpk3Rup3RBX4CjmLq
- [BUG] Unexpected session model tokens usage for opus
- Security blocked trying to defend my system!
- [Bug Report] POWER USER PISSED OFF - CLAUDE IS BROKEN AND GETTING WORSE NOT BETTER - HELP PLEASE!
- [Bug] VSCode extension ignores model selection and reverts to Opus despite Sonnet configuration
- [BUG] Claude Code fails to follow established process rules despite repeated correction
- [BUG] Claude Code ignored explicit file write prohibition in ~/.claude/instructions.md
- Silent model downgrade from Opus to Sonnet after re-authentication (Max subscriber)
- [Bug] Anthropic API not respecting MAX_THINKING_TOKENS limit, exceeds configured threshold
- [Bug] Model removes French accents from text output
- [Bug] Misleading API Model Limit Handling for Claude Models
- [BUG] Python SDK: `system_prompt` parameter causes `model` parameter to be ignored
- [Bug] Sonnet 4 model degradation: Generating Python Scripts with Syntax Errors
- [Bug][aup] Safety filter blocked frustrated exclamation mid-troubleshooting during profile migration task (req_011CcjQZaS6pYuMWY6EESxNE)
- [Bug][cyber] Debugging an H.264 video capture/extraction pipeline against known-good decoder output blocked (req_011CcY9KbFHf4xez9uNGQyAH)
- [Bug] Tool-call markup leaks into assistant text after large context accumulation
- [Bug][aup] ClAudit false-positive in [USER] — req_011CcPfKYfUStpKMjL3ppuVe
- [Bug][aup] Reviewing and hardening IAM roles/policies on own cloud tenant blocked as policy violation (req_011CcPdudXdWL8aMKMCfewMQ)
- [Bug] Parallel agent spawning causes excessive token consumption before crashing or hitting limits
- [Bug] Anthropic API Error: Over-aggressive content filtering blocking legitimate code review requests
- [Bug] Anthropic API Error: Overly Restrictive Content Policy on Abstract Mathematical Simulations
- Actions on Gromacs files triggers false "Usage Policy Violation" error in Opus
- [Bug] Advisor narrates but then fails to make tool calls
- Opus 4.7 tool-use compliance cliff between 24th-26th May — instrumented evidence
- [BUG] Confident hallucinations and gaslighting on user-described systems — getting worse since Opus 4.7
- [BUG] Agent makes unasked changes, adds assumptions, breaks working code
- [Bug] Claude ignores CLAUDE.md virtual environment directives despite loading file at session start
- Content filtering blocks legitimate development work mid-session, causing work loss
- [BUG] Task tool subagent model not displayed in TUI for opus and default cases
- Drastically reduced quality from Sonnet 4.5 and Claude Code ~Dec 26
- [Bug] Anthropic API Error: Cannot modify thinking blocks in compact operation
- Hallucination and Overconfidence in Code Generation Responses
- Issue Title: `find` Command Fails: Incorrect `-exec` Argument Syntax
- [Bug] Anthropic API Error: Content Flagged by Safety Filters on Legitimate Code Tasks
- [Bug] Anthropic API Error: Message flagged by safeguards policy
- [Bug] Fable model incorrectly triggers safeguards on scientific analysis discussion
- [Bug][cyber] Safety block halted authorized reverse-engineering of my own drone's auth protocol, mid-secret-re (req_011CcnReWU19UF1bg7REcCRy)
- [Bug][cyber] False-positive block on rooted-device SELinux/dev-mode setup for the user's own Android phone (req_011CcmzvbxgnDzowZjuRR3rJ)
- [Bug][aup] Safeguards block fired on resuming a defensive vulnerability scrub of user's own healthcare web app (req_011CccLKmK4wfjUQeJVPJ82C)
- [Bug][cyber] ClAudit false-positive in com — req_011CccPbdpfStasyx1ehijPs
- [Bug][aup] Blocks safe, validated edit to own web-server config and follow-up security audit (req_011CccLcALMtWAdVqZWvPu3G)
- Spurious token "court" emitted between tool calls breaks tool-call parsing and aborts the turn
- [Bug][cyber] Safety block prevents writing or reviewing code that reads drone telemetry sensor data (req_011CcYR3zvFDq6wKPwZEMFur)
- [Bug][cyber] Safety block interrupts legitimate security vulnerability analysis workflow (req_011CcY1o5TFLzNFaV7gPUFGo)
- [Bug][cyber] Blocked while reviewing a decompiled white-box AES class for app interop work (req_011CcV5MVf2AARSr2duriXsi)
- [Bug][cyber] Safety block interrupts legitimate open-source drone ground station development mid-session (req_011CcUWvRreLL6jx3hQHP7ob)
- [Bug][cyber] Safety block stops legitimate consumer drone firmware analysis via USB protocol inspection (req_011CcUBFqhJcvDBDuEgVegwF)
- [Bug][cyber] Consumer drone firmware download and version diff analysis for owned device wrongly blocked (req_011CcUAZiJCHXvDprqjwtBPp)
- [Bug][aup] Building a log-parsing tool that auto-files GitHub issues for safety false-positive refusals got bl (req_011CcPLnkSBdTjTa8KTcV63s)
- [BUG]
- [Bug] Tool-use blocks corrupted with prepended "count" token, causing parse failures
- [BUG] Resumed background subagent/teammate loses its spawn-time model override and reverts to parent model
- [Feature Request] Remove unnecessary verbal disclaimers.. All of them. They waste energy, time, tokens and money. Stop the rot.
- [BUG] Opus deleted entire repository without realizing it (corrupt tool call?)
- [Bug] Claude responding in Japanese when prompted in Chinese
- [BUG] Opus 4.7 subagents enter generation loop on post-task finalization/return text (work completes, agent doesn't exit)
- [BUG]
- [MODEL] Repeated unresearched API changes broke production business system; explicit welfare instructions ignored throughout
- I cannot generate a GitHub issue title from this text because it does not describe a technical bug or feature request. The message appears to be personal feedback or praise about Claude's performance, but does not identify a specific problem, error, or fea
- Skill subprocess inherits 1M context flag but switches model, fails with 'Extra usage is required for 1M context'
- Subagent `model:` frontmatter pin silently ignored in 2.1.117
- WebSearch tool does not support query operators, but agents construct queries as if it does — wasting tokens and producing degraded results
- [BUG] Claude did not follow instructions from its MD file multiple times
- [BUG] model changes automatically during one task
- [Bug] Anthropic API Error: tool_reference blocks unsupported on claude-haiku-4-5-20251001
- [BUG] Claude code fails when I set a model to use for custom commands
- Invalid OpenAI Model Name in API Request
- [Bug] Claude Code makes unsubstantiated guesses despite documented instructions and memory
- [Bug][cyber] Safety block halted routine memory-dump analysis of a mobile game process for reverse engineering (req_011CcnXGX2KkGqEv89H61eRm)
- [Bug][cyber] Safety filter blocked continued work on a drone auth plugin (RSA key-based protocol handshake cod (req_011CcnWhgbceAtzYUrYPwG71)
- [Bug][cyber] Safeguard blocked a routine "finish the task" continuation message with no security content (req_011CcnRnv4YvFFpaz45AjrWk)
- [Bug][cyber] False positive block during memory scan/watchpoint debugging tool development and testing (req_011CcnFJZScjXWEQuM7jUVN4)
- [Bug][cyber] False positive: safeguard blocked routine cloud IAM policy review and permission-scoping work (req_011Ccm9AiYP35wZYvf5jbWtw)
- [Bug][cyber] Safety block fired on routine request to audit repo for leaked secrets before making it public (req_011CckvjkmVEng3NE2wBmhGv)
- [Bug][aup] Fable 5 blocks legitimate user migration work when searching for admin auth recovery method (req_011CcjWcoWh6TZ5tN11bisQ1)
- [Bug][aup] False positive: safety filter blocked routine request to check GUI/UI status of devices after frust (req_011CcjQqFggPxZ1ryYKCAAFM)
- CONTINUE[Bug] Safeguard triggered incorrectly without unauthorized security request
- Quality regression: no cross-session context retention, repeated failures on same task
- [Bug][cyber] False cyber block while building drone flight UI with live video, telemetry, and connection-statu (req_011CcYLSQEyq2CeAsBAgK4Pz)
- [Bug][cyber] Correcting a too-wide on-screen video aspect ratio via ffmpeg parameter tweaks (req_011CcYKHENYHwr9vFJzc6eDX)
- [Bug][cyber] Building drone flight UI with live video + telemetry connection-status HUD blocked (req_011CcYJgUdojjgA9CLvAaxrz)
- [Bug][cyber] Cyber filter blocked building a drone flight UI with live video feed and telemetry HUD (req_011CcYJWT9EjGMeSaqybt7TC)
- [Bug][cyber] Safety block halted routine GUI work on a drone telemetry/video ground-station HUD (req_011CcYJUCfHYwm1Zzc8uwssh)
- [Bug][cyber] H.264 decoder debug blocked while inspecting CABAC corruption and unknown NAL types (req_011CcWya3V5wtjqdrbgxJhpd)
- [Bug][cyber] Safety block halted H.264 NAL-unit forensics: diagnosing CABAC corruption from suspicious NAL typ (req_011CcWyQ9hrc9xzACDwmp8mY)
- [Bug][cyber] Safety block interrupts APK unpacking/DEX decryption key analysis mid-session (req_011CcWUiMYd6DwbhzJSthUCh)
- [Bug][cyber] Safety block persisting wireless ADB debug across reboots on custom Android ROM (req_011CcWCLGsNCNEadxuojYP9Z)
- [Bug][cyber] Cloud IAM audit/review work incorrectly flagged as unsafe cybersecurity content (req_011CcVvH7oDPuj7ZVtVbfnJy)
- [Bug][cyber] Safety block on consumer drone firmware downgrade feasibility question (req_011CcVtCm2xFhDbukwuYU5eB)
- [Bug][cyber] False block halts authorized defensive security automation app development (req_011CcV5RdTBsufs4i5UgJgz3)
- [Bug][cyber] Safety block halted reverse-engineering of a vendor SDK's white-box AES decrypt routine (req_011CcV5NTXoGuPu7Ju6gDkxg)