Repository navigation
Codex CLI: broad slowdown in task execution and response generation, with slow compaction #49287
Description
What version of Codex CLI is running?
0.159.0, confirmed by /status during the retests and by a current codex --version check. This is also the latest stable GitHub release when checked for this report. I updated at OpenAI Support's request. Historical runtime versions have not been independently verified.
What subscription do you have?
ChatGPT Pro. /status showed approximately 80% of the weekly allowance remaining.
Which model were you using?
Selected model: gpt-6-sol; reasoning effort: Extra High (xhigh); recorded thread setting: service_tier=default. Authentication: ChatGPT sign-in; provider: OpenAI.
What platform is your computer?
Windows / WSL2; the test repository is on the Linux filesystem inside the home directory. I use Codex CLI and have never used Codex in VS Code.
Current uname -mprs output, collected after the retests:
Linux 6.18.33.2-microsoft-standard-WSL2 x86_64 unknown
What terminal emulator and version are you using (if applicable)?
A WSL terminal. The exact terminal emulator/version was not captured.
Codex doctor report
Collected at 2026-09-29T15:33:21Z, after the retests, from a different current workspace. It reports CLI 0.159.0 and overall status ok; active thread overrides were not inspected. This is a current diagnostic snapshot, not a reconstruction of the tested thread's exact runtime state. Local paths, project identifiers and an unrelated historical source inventory have been redacted.
Reviewed doctor report
{
"schemaVersion": 1,
"generatedAt": "1790696001s since unix epoch",
"overallStatus": "ok",
"codexVersion": "0.159.0",
"checks": {
"app_server.status": {
"id": "app_server.status",
"category": "app-server",
"status": "ok",
"summary": "background server is running",
"details": {
"app-server version": "0.159.0",
"control socket": "[redacted local path or project identifier]",
"daemon state dir": "[redacted local path or project identifier]",
"dedicated pid file": "[redacted local path or project identifier]",
"dedicated update-loop pid file": "[redacted local path or project identifier]",
"mode": "ephemeral",
"pid file": "[redacted local path or project identifier]",
"settings": "[redacted local path or project identifier]",
"status": "running",
"update-loop pid file": "[redacted local path or project identifier]"
},
"remediation": null,
"durationMs": 5
},
"auth.credentials": {
"id": "auth.credentials",
"category": "auth",
"status": "ok",
"summary": "auth is configured",
"details": {
"auth file": "[redacted local path or project identifier]",
"auth storage mode": "File",
"stored API key": "false",
"stored ChatGPT tokens": "true",
"stored agent identity": "false",
"stored auth mode": "chatgpt"
},
"remediation": null,
"durationMs": 0
},
"config.load": {
"id": "config.load",
"category": "config",
"status": "ok",
"summary": "config loaded",
"details": {
"CODEX_HOME": "[redacted local path or project identifier]",
"active thread overrides": "not inspected",
"config.toml": "[redacted local path or project identifier]",
"config.toml parse": "ok",
"configuration load ms": "19",
"configuration scope": "invocation config, including cloud-managed policy",
"cwd": "[redacted local path or project identifier]",
"enabled feature flags": "analytics_plan_history, daemon_auto_start, shell_tool, view_image, sleep_tool, unified_exec, unified_exec_tty, unified_exec_zsh_fork, shell_snapshot, content_item_kinds, code_mode_host, terminal_resize_reflow, sqlite, memories, write_stdin_approval, hooks, enable_request_compression, unbounded_connection_retries, worktrees, system_proxy_fallback, multi_agent, apps, tool_search_always_defer_mcp_tools, tool_suggest, plugins, in_app_browser, in_app_chat, in_app_dictation, in_app_local_automation, in_app_updates, browser_use, browser_use_full_cdp_access, browser_use_external, computer_use, remote_plugin, plugin_sharing, image_generation, resize_all_images, item_ids, skill_mcp_dependency_install, skill_search, mentions_v2, steer, guardian_approval, guardian_reuse_parent_compaction, goals, collaboration_modes, tool_call_mcp_elicitation, auth_elicitation, fast_mode, realtime_conversation, tui_app_server, compaction_image_budget, workspace_dependencies",
"feature flag overrides": "analytics_plan_history=true, memories=true",
"feature flags enabled": "54",
"log dir": "[redacted local path or project identifier]",
"mcp servers": "0",
"model": "gpt-6-sol",
"model provider": "openai",
"sqlite home": "[redacted local path or project identifier]"
},
"remediation": null,
"durationMs": 0
},
"git.environment": {
"id": "git.environment",
"category": "git",
"status": "ok",
"summary": "git executable found; execution not verified",
"details": {
".git entry": "directory",
"PATH git #1": "[redacted local path]",
"PATH git #2": "[redacted local path]",
"PATH git entries": "2",
"git execution": "not inspected (PATH helpers are not executed)",
"repo detected": "true",
"repo root": "[redacted local path or project identifier]",
"selected git": "[redacted local path]"
},
"remediation": null,
"durationMs": 113
},
"installation": {
"id": "installation",
"category": "install",
"status": "ok",
"summary": "installation looks consistent",
"details": {
"PATH codex #1": "[redacted local path or project identifier]",
"current executable": "[redacted local path or project identifier]",
"install context": "[redacted local path or project identifier]",
"managed by Vite+": "false",
"managed by bun": "false",
"managed by npm": "false",
"managed by pnpm": "false",
"managed package root": "not set"
},
"remediation": null,
"durationMs": 157
},
"mcp.config": {
"id": "mcp.config",
"category": "mcp",
"status": "ok",
"summary": "no MCP servers configured",
"details": {},
"remediation": null,
"durationMs": 0
},
"network.env": {
"id": "network.env",
"category": "network",
"status": "ok",
"summary": "network-related environment looks readable",
"details": {
"managed proxy": "not configured",
"proxy env vars": "none",
"respect system proxy": "disabled"
},
"remediation": null,
"durationMs": 0
},
"network.provider_reachability": {
"id": "network.provider_reachability",
"category": "reachability",
"status": "ok",
"summary": "active provider endpoints are reachable over HTTP",
"details": {
"ChatGPT inference URL": "https://chatgpt.com/backend-api/<redacted> reachable (HTTP 405)",
"reachability mode": "ChatGPT auth"
},
"remediation": null,
"durationMs": 912
},
"network.websocket_reachability": {
"id": "network.websocket_reachability",
"category": "websocket",
"status": "ok",
"summary": "Responses WebSocket handshake succeeded",
"details": {
"DNS": "1 IPv4, 0 IPv6, first IPv4",
"auth mode": "chatgpt",
"connect timeout": "15000 ms",
"endpoint": "wss://chatgpt.com/backend-api/<redacted>",
"handshake result": "HTTP 101 Switching Protocols",
"model provider": "openai",
"provider name": "OpenAI",
"proxy env vars": "none",
"reasoning header": "false",
"server model present": "false",
"supports websockets": "true",
"wire API": "responses"
},
"remediation": null,
"durationMs": 1515
},
"runtime.provenance": {
"id": "runtime.provenance",
"category": "runtime",
"status": "ok",
"summary": "running standalone on linux-x86_64",
"details": {
"commit": "unknown",
"current executable": "[redacted local path or project identifier]",
"install method": "[redacted local path or project identifier]",
"platform": "linux-x86_64",
"version": "0.159.0"
},
"remediation": null,
"durationMs": 0
},
"runtime.search": {
"id": "runtime.search",
"category": "search",
"status": "ok",
"summary": "search command found (bundled); execution not verified",
"details": {
"search command": "[redacted local path or project identifier]",
"search command readiness": "file exists",
"search provider": "bundled"
},
"remediation": null,
"durationMs": 0
},
"sandbox.filesystem_paths": {
"id": "sandbox.filesystem_paths",
"category": "sandbox",
"status": "ok",
"summary": "no explicit filesystem paths to probe",
"details": {},
"remediation": null,
"durationMs": 0
},
"sandbox.helpers": {
"id": "sandbox.helpers",
"category": "sandbox",
"status": "ok",
"summary": "sandbox configuration is readable",
"details": {
"approval policy": "OnRequest",
"codex-linux-sandbox helper": "[redacted local path or project identifier]",
"denied-read glob rules": "0",
"denied-read rules": "0",
"execve wrapper helper": "[redacted local path or project identifier]",
"filesystem sandbox": "restricted",
"glob scan max depth": "unbounded",
"managed filesystem source": "none",
"network sandbox": "restricted"
},
"remediation": null,
"durationMs": 0
},
"security.endpoint": {
"id": "security.endpoint",
"category": "security",
"status": "ok",
"summary": "endpoint protection is not inspected on this platform",
"details": {
"endpoint products": "not inspected on this platform"
},
"remediation": null,
"durationMs": 0
},
"state.paths": {
"id": "state.paths",
"category": "state",
"status": "ok",
"summary": "state paths and databases are inspectable",
"details": {
"CODEX_HOME": "[redacted local path or project identifier]",
"active rollout files": "1742 files, 5675081108 total bytes, 3257796 average bytes",
"archived rollout files": "0 files, 0 total bytes, 0 average bytes",
"goals DB": "[redacted local path or project identifier]",
"goals DB integrity": "ok",
"log DB": "[redacted local path or project identifier]",
"log DB integrity": "ok",
"log dir": "[redacted local path or project identifier]",
"memories DB": "[redacted local path or project identifier]",
"memories DB integrity": "ok",
"memories v2 DB": "[redacted local path or project identifier]",
"memories v2 DB integrity": "skipped (missing)",
"queue DB": "[redacted local path or project identifier]",
"queue DB integrity": "ok",
"sqlite home": "[redacted local path or project identifier]",
"standalone release cache": "[redacted local path or project identifier]",
"state DB": "[redacted local path or project identifier]",
"state DB integrity": "ok",
"thread history DB": "[redacted local path or project identifier]",
"thread history DB integrity": "ok"
},
"remediation": null,
"durationMs": 20567
},
"state.rollout_db_parity": {
"id": "state.rollout_db_parity",
"category": "threads",
"status": "ok",
"summary": "rollout files and state DB thread inventory agree",
"details": {
"default model provider": "openai",
"rollout DB active files": "1742",
"rollout DB active rows": "1742",
"rollout DB archive mismatches": "0",
"rollout DB archived files": "0",
"rollout DB archived rows": "0",
"rollout DB duplicate DB paths": "0",
"rollout DB duplicate rollout thread ids": "0",
"rollout DB malformed file names": "0",
"rollout DB missing active rows": "0",
"rollout DB missing archived rows": "0",
"rollout DB model providers": "openai=1742",
"rollout DB rows": "1742",
"rollout DB scan cap reached": "false",
"rollout DB scan errors": "0",
"rollout DB sources": "[redacted unrelated historical source inventory; not a runtime-client measurement]",
"rollout DB stale rows": "0"
},
"remediation": null,
"durationMs": 732
},
"system.disk": {
"id": "system.disk",
"category": "disk",
"status": "ok",
"summary": "sufficient free disk space (936.1 GiB)",
"details": {
"CODEX_HOME available": "936.1 GiB",
"failure threshold": "1.0 GiB",
"warning threshold": "5.0 GiB",
"worktree available": "936.1 GiB"
},
"remediation": null,
"durationMs": 0
},
"system.environment": {
"id": "system.environment",
"category": "system",
"status": "ok",
"summary": "OS language C",
"details": {
"EDITOR": "not set",
"GH_PAGER": "set",
"GIT_PAGER": "set",
"LANG": "C.UTF-8",
"LC_ALL": "C.UTF-8",
"LC_CTYPE": "C.UTF-8",
"PAGER": "set",
"VISUAL": "not set",
"os": "Ubuntu 26.4.0 (resolute) [64-bit]",
"os language": "C",
"os type": "Ubuntu",
"os version": "26.4.0"
},
"remediation": null,
"durationMs": 7
},
"terminal.env": {
"id": "terminal.env",
"category": "terminal",
"status": "ok",
"summary": "terminal metadata was detected",
"details": {
"COLORTERM": "present",
"DISPLAY": "present",
"NO_COLOR": "1",
"TERM": "xterm-256color",
"WAYLAND_DISPLAY": "present",
"WSL_DISTRO_NAME": "present",
"WSL_INTEROP": "present",
"color output": "disabled (NO_COLOR)",
"effective locale": "C.UTF-8",
"stderr is terminal": "false",
"stdin is terminal": "false",
"stdout is terminal": "false",
"terminal": "unknown",
"terminal size": "80x24"
},
"remediation": null,
"durationMs": 1
},
"terminal.title": {
"id": "terminal.title",
"category": "title",
"status": "ok",
"summary": "terminal title default",
"details": {
"terminal title activity": "true",
"terminal title items": "activity, project-name",
"terminal title project source": "git repo root",
"terminal title project value": "[redacted local path or project identifier]",
"terminal title source": "default"
},
"remediation": null,
"durationMs": 0
},
"updates.status": {
"id": "updates.status",
"category": "updates",
"status": "ok",
"summary": "update configuration is locally consistent",
"details": {
"cached latest version": "0.158.0",
"check for update on startup": "true",
"last checked at": "2026-09-29T05:09:43.375380566Z",
"latest version": "0.159.0",
"latest version status": "current version is not older",
"update action": "standalone installer",
"version cache": "[redacted local path or project identifier]"
},
"remediation": null,
"durationMs": 1891
}
}
}What issue are you seeing?
I am reporting a broad slowdown in everyday Codex work, including task execution and response generation. My usual Codex work now feels much slower throughout a task. Similar work that previously took about 5 minutes now often takes about 20 minutes. I also perceive substantially lower output throughput. These are observations from normal use, not a controlled replay of the same tasks or a measured historical decoding-TPS benchmark.
Context compaction is one affected operation and provides the clearest retained timing evidence. It previously took roughly one minute in my experience and now often takes 3–4 minutes. Local compaction records show longer intervals by September 28; the exact onset and first affected runtime version of the overall slowdown have not been established.
Verification performed after updating to CLI 0.159.0 at OpenAI Support's request:
| Test | Duration | Outcome / timing source |
|---|---|---|
| Ordinary response in a fresh CLI thread: 100 fixed numbered lines | 45.208 s total; 11.747 s recorded time to first token | All 100 lines; zero tool calls; local session records |
| Identical prompt in a fresh ChatGPT Work web chat | About 32 s total; first visible line about 5 s | All 100 lines; manual timing |
Existing long CLI thread, manual /compact |
187.277 s | Success; about 136K / 258K context used |
Fresh CLI thread, manual /compact after the fixed-output prompt |
14.521 s | Success; approximately 4% context used according to the interface |
Both CLI tests used the same version, repository, selected model and reasoning effort. The fresh CLI thread was started with --no-daemon. The browser comparison used ChatGPT Work on the web, not a separately verified Codex cloud task. Browser input tokens, system context and routing are unknown, and its first-visible-line timing differs from the CLI's time-to-first-token metric.
The ordinary CLI response recorded 21,064 input tokens, including 12,800 cached tokens, and 1,244 output tokens, including 138 reasoning tokens. The large compaction recorded 136,548 input tokens, 6,912 cached tokens and 5,119 output tokens. The ordinary request's input count is a separate measurement from the user-reported 4% context indicator. These turn timings do not measure pure token decoding TPS or provide a controlled historical baseline.
The long and fresh compaction durations differ substantially, but that comparison alone does not establish the cause of the broader slowdown in ordinary Codex work. A faster small-thread /compact does not resolve my overall performance complaint. Please investigate ordinary response/task latency as well as compaction.
For historical context, the original retained compaction extract shows:
| UTC date | Samples | Median duration |
|---|---|---|
| 2026-09-27 | 3 | 50.424 s |
| 2026-09-28 | 22 | 190.348 s |
| 2026-09-29 | 4 | 188.523 s |
These cohorts involve different tasks and contexts, end before the manual retests, and are observational data. Historical runtime versions have not been independently verified. The root cause remains unresolved; an account restriction has not been established.
What steps can reproduce the bug?
Uploaded thread: 01a0e606-748c-71c0-822f-b9cd8bf7efaf
Observed workflow and diagnostic checks; this is not a deterministic minimal reproducer for every account:
- Use Codex CLI with ChatGPT sign-in and GPT-6 Sol / Extra High. The retests below used 0.159.0.
- Perform normal coding work and observe overall task completion time and response/output speed. My reported change is approximately 5 minutes to 20 minutes for similar everyday work; I do not have a controlled old-versus-new task replay.
- Start a fresh CLI thread in the same repository with the same model and effort. Send the fixed prompt below and record ordinary response latency. This produced the 45.208-second response with zero tool calls.
- Run
/compactin that fresh thread and record its duration. Separately resume an existing long coding thread, record/status, then run/compact. This produced the 14.521-second and 187.277-second samples. - For an additional browser comparison, send the identical prompt in a fresh ChatGPT Work web chat with GPT-6 Sol / Extra High. The approximate timings are listed above.
The exact synthetic test prompt was:
不要读取文件或调用工具。
输出100行，编号从001到100。
每行编号后都写：The quick brown fox jumps over the lazy dog.
只输出这些行。
What is the expected behavior?
Codex should retain its earlier responsiveness for comparable everyday work, including ordinary responses and task execution. Please investigate the overall slowdown using both the ordinary-response and compaction traces below, and distinguish queuing, input processing, model generation, transport/retries and client delays where backend telemetry permits.
Please also clarify whether approximately three minutes is expected for this compaction workload. Even if context size explains that particular interval, the reported slowdown across ordinary work still needs investigation. Clear progress/timing information would help users understand where the time is spent.
Additional information
Ordinary CLI response trace (100 fixed lines, no tool calls):
- Thread/session ID:
01a0ed76-b2ef-70d2-a489-06d1382197aa - Response ID:
resp_077ac309f514e4f5016abbc9cd3f4887d08927fff1a33f8ca1 - UTC turn window:
2026-09-29T14:23:05.292Z–2026-09-29T14:23:50.490Z - Recorded turn duration:
45208 ms; recordedtime_to_first_token_ms=11747.
Large-thread compaction trace:
- Thread/session ID:
01a0e606-748c-71c0-822f-b9cd8bf7efaf - Response ID:
resp_0248d1da218a50c6016abbba1d2ff487d094a7f7e603f88575 - Compaction item ID:
01a0ed4e-f4dc-7050-b37a-3ab2d8c38b23 - UTC window:
2026-09-29T13:16:09.820Z–2026-09-29T13:19:17.097Z
Fresh-thread compaction trace:
- Thread/session ID:
01a0ed76-b2ef-70d2-a489-06d1382197aa - Compaction item ID:
01a0ed8f-c2ec-78d0-8bf6-4ba8191a87c0 - UTC window:
2026-09-29T14:26:56.876Z–2026-09-29T14:27:11.397Z
Earlier browser/authentication observations:
Before noticing the slowdown, I experienced two episodes of persistent ChatGPT web errors with many HTTP 403 requests in the browser console. Refreshing did not resolve them, and I signed out and back in approximately three times while troubleshooting. Exact timestamps were not captured. Those browser errors have resolved, but my overall Codex performance complaint remains. I have not seen suspicious-login warnings, security-risk banners or notices of additional processing, and I have not performed an authorized cybersecurity audit. These observations do not establish that the account was flagged or that the browser errors caused the slowdown. Please check for account/session-specific performance restrictions only if backend evidence supports that investigation.
Feedback submission:
Two /feedback attempts including logs failed. Local diagnostics for each recorded a feedback event uploaded to Sentry, six successful attachment uploads, and attachments_failed=true, with elapsed times of 300001 and 300000 ms. The TUI displayed Failed to upload feedback: feedback/upload failed in TUI. The main rollout measured 58,989,815 bytes during the investigation; the failed attachment and underlying failure cause are unknown.
A third attempt excluding logs immediately displayed Feedback recorded (no logs) and the existing thread ID above. It directed me to this prefilled CLI Bug form. Complete session-log upload is unverified; no separate feedback receipt ID was printed. The upload failure is a separate observation, not an established cause of the performance slowdown.
OpenAI Support has confirmed that the previously provided trace identifiers are sufficient for backend correlation and has escalated the private conversation to a support specialist. No investigation result has been provided yet.
Related open reports: #47656 concerns broader task latency and includes a CLI report; #45433 concerns desktop automatic-compaction latency. My report provides current CLI ordinary-response and manual-compaction measurements, plus a broader user-observed task slowdown. A shared root cause has not been established; please consolidate with a related tracker if appropriate. Historical compaction report #24372 is closed.
Activity
[-]CLI 0.159.0: /compact takes 187 s at ~136K context vs 14.5 s in a fresh thread[/-][+]Codex CLI: broad slowdown in task execution and response generation, with slow compaction[/+]KevinTokano commentedon Oct 1, 2026
I am experiencing the same severe Codex CLI performance degradation since September 29.
Performance is approximately 4x–9x slower than the previous normal baseline, depending on the session and time. This affects normal model operation across multiple models, not just compaction.
As of October 1, the degradation has become even worse and Codex CLI is effectively unusable for normal development work.
I submitted
/feedbackfrom an affected session. The diagnostic upload failed withfeedback/upload failed in TUI, but the feedback itself was successfully recorded without logs.Thread ID:
01a0f7e7-4c59-7a21-902f-8cd9e9a9b041
stephanemagnenat commentedon Oct 3, 2026
I am experiencing a similar problem under Linux with latest Codex CLI version 0.160.0 and Sol 6.1 medium effort normal speed.
heyarny commentedon Oct 3, 2026
I’ve been having the same problems since the new subscriptions were announced. That can’t be a coincidence.