Model behaviour issues on macOS
Issues the maintainers labelled both platform:macos and area:model.
2,520 issues · 662 open · 1,858 resolved (74%) · first seen May 20, 2025
Is this getting better or worse?
This class of problem is still growing. 1510 new reports in the last 90 days vs 654 in the 90 before — +131%. The open backlog peaked at 681 in 2026-07 and sits at 662 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
1,908 of these reports name the Claude Code build they were running, spanning 181 releases. Heaviest reporters:
- Claude Code v2.1.198
- Claude Code v2.1.170
- Claude Code v2.1.201
- Claude Code v2.1.207
- Claude Code v2.1.197
- Claude Code v2.1.199
- Claude Code v2.1.212
- Claude Code v2.1.211
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 35 days across 1,858 closures. Of the 1,858 closures with a recorded reason, 13% were closed as completed and 1,611 as not-planned or duplicate. The most recent completed fix landed in 2026-08.
Workarounds reported by the community
208 of these issues post a workaround someone says worked. The highest-engagement ones:
- [Bug] Anthropic API Error: Model's tool call could not be parsed (retry also failed)
- This issue makes claude code unusable... Quick workaround:
Found in the thread of #62123 · still open - [BUG] Opus 4.6 removed from Claude Desktop Code tab model picker after Opus 4.7 release
- …he option to use Opus 4.6 but not the "Coding" tab? Anyone has a workaround for the Desktop App?
Found in the thread of #49689 · resolved - Model setting changes globally across all sessions (regression)
- …ly in , so changing it in one session affects all sessions. Workaround 1 — Use project-level model override: Set the model per-project in : Project-level settings override the global setting, so each project can have its…
Found in the thread of #20745 · resolved - [BUG] Claude clobbers correct typographic marks and can't be told otherwise
- …ization code - Any scenario where Unicode precision matters Workaround: Use Unicode escape sequences (e.g., ) instead of literal Unicode characters. ------------ I've not yet tested the workaround, but I imagine it'd wor…
Found in the thread of #1599 · still open - [BUG] Claude Code defaults to 1M context on fresh session with no workaround on Pro plan
- …ars immediately without any prompt or file loaded Attempted workarounds — none resolved it: - — same error - — same error - → with Pro credentials only — same error Expected behavior: Standard context window used by defa…
Found in the description of #62063 · resolved
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 1–100
Ranked by community engagement (reactions weighted over comments). Listing the top 300 of 2,520 matches; the statistics above cover all 2,520.
- [Bug] Anthropic API Error: Model's tool call could not be parsed (retry also failed)
- [BUG] Opus 4.6 removed from Claude Desktop Code tab model picker after Opus 4.7 release
- Model setting changes globally across all sessions (regression)
- [BUG] Claude clobbers correct typographic marks and can't be told otherwise
- [BUG] Claude Code defaults to 1M context on fresh session with no workaround on Pro plan
- [BUG] Abnormal / inflated rate limit / session usage
- [BUG] Bash tool calls emitted as raw <invoke> text instead of executing
- Advisor tool returns "unavailable" on claude-fable-5 whenever transcript exceeds ~100K tokens
- [BUG] Fable 5 prompts 'usage credits required' on Max plan from 2026-07-20, the day it became standard on Max (saved model claude-fable-5[1m])
- Model behavior: /goal Stop-hook directive cited as authorization for unrequested actions; absence-from-search treated as evidence of absence; structure-as-substance under pushback
- [Bug] [URGENT] Claude Opus 4.8/5.0 reasoning degradation, speed and performance regression
- Plan mode: model made file edits despite plan mode being active
- [BUG] Opus 4.7 mixes legacy XML tool-use format into JSON tool calls on longer payloads
- [FEATURE] Model switching for existing Cowork tasks within Projects
- Tool calls intermittently emitted as literal text (stray 'court' + raw <invoke>) instead of executing — mostly Edit/Read (Opus, CLI)
- [Bug] Plan mode resurfaces old plans after session compaction with Claude 4.5 Opus
- [Bug] 1M context incorrectly requires extra usage on Max plan
- [Bug] /context command reports 200K max for Claude Opus 4.6 instead of 1M
- [BUG] German umlauts (ä, ö, ü) randomly replaced with ASCII substitutes (ae, oe, ue)
- Opus 4.6 (1M context) hits 'API Error: Rate limit reached' on Max subscription
- [Bug] Advisor with Fable 5 returns "advisor unavailable" error
- Sonnet 4.6 1M context unavailable on Max 20x plan - only Opus 1M appears in /model
- Opus 4.8 not selectable in CLI `/model` despite being available on account
- [BUG] Skill `model:` frontmatter triggers 429 "Extra usage required" on 2.1.76 when session uses opus[1m]
- [BUG] Armenian script generation broken - outputs "delays" instead of Armenian characters
- [BUG] LLM biased against accessible code
- Model fabricates tool output (and even a user instruction) when a parallel batch is partially cancelled / appears empty
- [BUG][Bedrock] Agent Teams teammates spawned with non-Bedrock model ID via --model flag, causing 400 errors
- [BUG] Missing AWS Bedrock US Inference Profile for Opus 4.5
- [BUG][bedrock] Opus 4.6 [1m] selection writes invalid model id → 400 "The provided model identifier is invalid"
- [BUG] Severe Performance Degradation in Claude Code v1.0.81 - Frequent Hallucinations and Instruction Following Failures
- [BUG] MCP FastMCP Pydantic model parameters serialized as JSON strings causing validation errors
- [BUG] There's an issue with the selected model (claude-fable-5). It may not exist or you may not have access to it. Run /model to pick a different model.
- Massive quality regression
- [BUG] REGRESSION: Cannot switch between Opus and Sonnet mid-thread in Cowork after April 15 desktop redesign
- Compaction fails with 'Conversation too long' at 48% context usage (Opus 4.6)
- [Bug] Fable 5 safeguard `[reasoning_extraction]` false-positives on a one-word greeting ("Hi")
- [BUG] opusplan setting doesn't switch to Opus during plan mode
- Custom Bedrock models not discoverable in /model command list
- [BUG] AUP/cyber-safeguard false positives on legitimate own-software hardening; one hit contaminates entire session
- [BUG] macOS v1.0.51: Claude Max users incorrectly treated as Pro users - Cannot access Opus model
- [BUG] Invalid env var CLAUDE_CODE_MAX_OUTPUT_TOKENS
- [Bug] Claude is being lazy - very very lazy
- [BUG] Opus 4.7 missing from Claude Code CLI /model selector (available in Desktop App)
- [BUG] Opus 4.7 thinking renders empty/unexpandable stub in VS Code extension
- [BUG] Prompt Caching Works on Claude 3.7 But Not On Claude 4 Family (AWS Bedrock)
- Agent Teams: Spawned teammates don't inherit model configuration from team lead
- [Bug] Opus Reasoning Degradation: Incorrect Assumptions and Poor Comprehension
- [BUG] Haiku 4.5 does not appear in model selector
- [BUG] : Cannot use AWS Bedrock with Claude Code. Getting API Error (429 Too many tokens)
- [BUG] Fable 5 safeguards repeatedly flag benign messages in a legitimate anti-fraud (T&S) session — forced fallback to Opus 4.8
- [BUG] Claude Opus 4 limit reached, now using Sonnet 4
- [Bug] Unexpected Model Switching Between Terminal Sessions
- Effort/model selection ("ultracode") reverts to "extra" when switching chats
- v2.1.158 (Opus 4.8): main session missing Grep/Glob, Bash storms, duplicate Bash commands, fabricated branches/commits/files, excessive shell text commands
- [Feature Request] Improve security research context recognition to differentiate authorized penetration testing from malicious intent
- [BUG] Thinking block content empty (thinking_len=0) since v2.1.69 — only encrypted signature returned
- [BUG] `opusplan` mode falls back to `sonnet 3.7` instead of `sonnet 4` for edits
- [BUG] Haiku 4.5 usage being counted the same as Sonnet 4.5 despite 3x lower cost
- [BUG] CLAUDE.md Instructions Systematically Ignored / Suspected Model Substitution - Opus 4.6
- [Bug] Stray token "count" injected before tool calls causes parse failure in Claude Code
- [Bug] Tool-use block markup intermittently corrupted with stray token prefix
- [BUG] Thinking mode 400 error on FIRST tool call in brand new conversation (Opus 4.5)
- [BUG] Claude Code changes model identifier mid-conversation on AWS Bedrock (eu to us region)
- [Bug] Anthropic API Error: Model claude-opus-4-7 temporarily unavailable, blocking safety classifier
- [BUG] This is not the opus 4.5 released some time ago
- [Bug] Opus 4.6 model performance degradation and effort configuration not persisting correctly
- [MODEL] Opus 4.6 does not read prompts properly — substitutes its own interpretation, causes rabbit holes
- [BUG] Claude Code does not respect agent model definition, uses wrong model.
- [BUG] Haiku 4.5 incorrectly limited to 8192 output tokens instead of 64K
- [BUG] ⏺ API Error: 400 tools.11.model: 'claude-opus-4-8' cannot be used as an advisor when the request model is 'claude-fable-5'.
- [BUG] Error 400 ""thinking.type.enabled" is not supported for this model" still exists in 2.1.113 when used with Opus 4.7 in Bedrock
- /model claude-opus-4-7 fails with "thinking.type.enabled" not supported error
- [BUG] Task tool `model` parameter returns 404 - Blocks cost-optimized workflows (10-30x unnecessary spending)
- [BUG] 2.1.68 Sends effort parameter when using custom model
- [Bug] Advisor "unavailable" = `advisorModel: fable` + any `tool_use` in the transcript; executor-irrelevant, `opus` advisor immune (supersedes #73923's ToolSearch cause)
- Opus claude-opus-4-8 emits unparseable tool calls at ~1.5% rate; opus-4-7 and sonnet-4-6 have 0% failure rate
- Max plan: sonnet[1m] requires extra usage but opus[1m] is included - inconsistent
- Agent Teams: Support per-teammate model configuration
- [BUG] API 400: thinking block ordering error during conversation with Task/Explore agent
- [BUG] Sonnet 1M context window unavailable on Max 20x subscription - regression from ~2 weeks ago
- /model shows Fable 5 as credits-only (€0.00) on Max plan, while /usage shows Fable covered by plan (5% of weekly limit)
- [Bug] Fable 5 writes user-facing output into hidden thinking instead of a text block — same as #66112, fixed for Opus but still reproduces on Fable
- [BUG] Token consumption with Fable 5 is not matching with its description.
- [Bug] Claude 4.8 Opus hallucinating tool outputs without execution
- [Bug] Anthropic API Error: Usage credits required for 1M context with 1M context disabled
- [Bug] Anthropic API Error: 400 Tool Concurrency Issues with Opus Model
- Model Mismatch: Max Plan Advertising Claude Opus 4 but Delivering Claude 3 Opus
- [Bug] Anthropic API Error: Incompatible model for tool advisor when using claude-fable-5
- [Bug] Desktop app (Code tab) 1.9659.4: Opus 4.8 "tool call could not be parsed (retry also failed)" still reproduces
- [BUG] Opus 4.8 in Claude Code declares work "verified" / "done" without running the canonical build — false-green regression vs. Opus 4.7
- [Bug] Claude Opus 4.8: Tool call parsing fails when extended thinking is triggered
- WebFetch can fabricate confident factual claims by misreading page boilerplate as live status
- [BUG] Claude Code continually ignores CLAUDE.MD file
- [BUG] 400 Input is too long for requested model with `global.anthropic.claude-sonnet-4-5-20250929-v1:0`
- [Bug] Fable 5 model safety measures blocking normal conversation content
- [BUG] Tool calls intermittently emitted as literal text (stray "count" + raw <invoke>) instead of executing — Opus 4.8, long Edit-heavy session
- [BUG] Opus 4.7 is trash. why is it severely so much worst than opus 4.6
- Advisor (Fable 5) returns "unavailable" whenever the transcript contains any prior tool call — one Bash(ls) suffices; Opus advisor immune
- Safety-classifier model switch (Fable 5 → Opus 4.8) fires on benign content and can't be overridden with /model