<- all tokdocs

Codex vs Claude Code: Fast Mode Costs, Email Triggers and Imports

By , founder of Mavric and Apso

View on TikTok ->

Codex Fast mode makes responses 1.5x faster, but it draws on your included subscription limits at 2.5x the Standard rate. That cost is the main thing a viral pitch to switch from Claude to ChatGPT leaves out. The same pitch praises ChatGPT's email-triggered automations, which OpenAI's docs say do not run in the ChatGPT desktop app the video demonstrates.

The claims come from a TikTok by @sabrina_ramonov that opens with "why everyone's switching back to ChatGPT." Several of her points check out, and others need corrections.

Did people switch "back" to ChatGPT from Claude Code?

The word "back" implies ChatGPT had the terminal coding agent first. It did not. Anthropic shipped Claude Code as a research preview on 24 February 2025 (Anthropic), and it reached general availability on 22 May 2025 with Claude 4. OpenAI released Codex CLI on 16 April 2025. Claude Code came first by about seven weeks.

The clip also gives no evidence for "everyone." It cites no survey, download count or usage figure.

How much does Codex Fast mode cost?

The video says Fast mode makes Codex "go 1.5x speed, even though it'll cost you more tokens, but this saves me time, so it's worth it."

OpenAI's speed documentation gives the number the voiceover leaves out. Fast mode "uses included subscription limits at 2.5x the Standard rate," and "purchased credits and Enterprise pay-as-you-go usage are billed at 2x the Standard rate." So you spend 2.5x the usage to get 1.5x the speed. The cost is a rate multiplier on your limits or credits, not a count of extra tokens.

Her on-screen blog text does warn that the cost is "oftentimes MORE than 1.5x," but a viewer who only listens never hears the multiplier.

Which models does Fast mode support?

The on-screen heading says "Fast mode raises speed by 1.5x on supported models." The OpenAI page is narrower. It lists "GPT-6.1 Sol, GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna, where available. For GPT-5.6 and GPT-5.5, the speed increase is 1.5x."

The 1.5x figure is stated only for GPT-5.6 and GPT-5.5. The demo shows "Model 5.6 Sol" selected, so it matches the documented case. OpenAI does not publish a speed multiplier for the GPT-6 models.

To turn it on, use /fast in the CLI, or set service_tier = "fast" plus [features].fast_mode = true in config.toml.

Do ChatGPT scheduled tasks support email triggers?

They do, with a limit the video skips. She says ChatGPT has a unified scheduler and "just release trigger-based workflow. For example, when an email comes in, then run this automation."

OpenAI's scheduled tasks page lists event triggers for Gmail ("New incoming messages, optionally filtered by sender or subject"), Slack and GitHub pull request activity. It then says event-triggered tasks are not available in the ChatGPT desktop app, the Codex CLI or the IDE extension. Those get time-based scheduling only.

Her own migration instructions are written for "the ChatGPT desktop app," one of the places the Gmail trigger does not run. OpenAI's changelog dates event triggers to 25 August 2026, nine days before the video, so "just release" is accurate.

Does Claude Code have trigger-based automation?

It does, and it shipped earlier. Anthropic's routines documentation describes three trigger types: "Scheduled," "API" through an authenticated HTTP POST to a per-routine /fire endpoint, and "GitHub," which runs "in response to repository events such as pull requests or releases." The API endpoint uses the beta header experimental-cc-routine-2026-04-01, which dates it to roughly April 2026, months before OpenAI's event triggers.

Her complaint that Claude's scheduling is fragmented is fair, and she understates it. Anthropic documents three separate surfaces: cloud routines, Desktop scheduled tasks that run on your machine, and /loop and in-session scheduling.

Can Claude generate images?

No, and the video is right about this. The FAQ on Anthropic's vision page says "Claude is an image understanding model only. It can interpret and analyze images, but it cannot generate, produce, edit, manipulate, or create images."

Codex can. OpenAI's image generation page says it uses gpt-image-2, invoked with $imagegen or by asking in chat. The video skips the cost: "Image generations use included limits 3-5x faster on average than similar turns without image generation, depending on image quality and size."

Is the Codex browser available everywhere?

It is not. The video's "CMD + T" shortcut is correct. OpenAI's browser documentation gives it as "Cmd+T (Ctrl+T on Windows and Linux)." The same page says "Browser isn't available in Codex CLI or the Codex IDE extension," so the feature belongs to the app.

How to migrate from Claude Code to Codex

Every step she reads out matches OpenAI's import guide. In the desktop app, open "Settings > Import." In the Codex CLI, run /import. The import covers:

  • Instruction files (AGENTS.md, settings.json, config.toml)
  • Skills, plugins and project folders
  • Project memories, MCP server configuration and hooks
  • Slash commands and subagents
  • Chat sessions from the last 30 days, capped at 50 chats in the CLI

The desktop app imports from Claude Code, Claude Cowork or Cursor. The CLI imports from Claude Code or Cursor. The changelog dates the feature to 11 August 2026.

Her skills-path detail is also right. OpenAI's skills documentation lists .agents/skills (both words plural) at the current directory, parent directory, repo root and home directory, plus /etc/codex/skills. Each skill is "a directory with a SKILL.md file plus optional scripts and references."

Bottom line

Codex Fast mode costs 2.5x usage on subscription limits for a documented 1.5x speedup on GPT-5.6 and GPT-5.5. Gmail event triggers exist but do not run in the ChatGPT desktop app, Codex CLI or IDE extension. Claude Code shipped before Codex and already had API and GitHub triggers. The video is correct that Claude cannot generate images and that the Claude Code import works as described.

Sources

Published September 3, 2026. Based on a TikTok by @sabrina_ramonov. Full video notes (markdown)