OpenAI Removes Codex 5-Hour Rate Limit: What 6M Users Mean for AI Coding Costs

Rate Limits Lifted Across All Paid Plans

OpenAI has temporarily removed the 5-hour rate limit on Codex for all Plus, Business, and Pro plan subscribers. Previously, Codex users hit a hard cap after a certain number of requests within a rolling 5-hour window, forcing developers to either wait or switch to direct API calls at higher cost. The removal means unlimited Codex usage for the duration of the promotional period.

This follows Anthropic's recent decision to extend the free usage window for Claude Fable 5, signaling an industry-wide push to maximize adoption during what appears to be a critical land-grab phase in AI coding tools. Both companies are betting that getting developers locked into workflows now will pay off when limits are eventually reimposed or pricing tiers shift.

GPT-5.6 Sol Efficiency Gains

Alongside the rate limit removal, OpenAI is pushing GPT-5.6 Sol efficiency optimizations that reduce token consumption for equivalent coding outputs. Sol, priced at $5/$30 per million tokens (input/output), is positioned as the workhorse model for coding tasks — faster than the full GPT-5.6 Terra ($2.50/$15) for most completions while maintaining high code quality.

The efficiency improvements work by reducing the number of tokens Sol generates to accomplish the same task. OpenAI reports that common coding operations — function generation, refactoring, test writing — now consume 15-25% fewer output tokens compared to the same model three months ago. For teams spending $500-2,000/month on API coding, this translates to $75-500 in monthly savings without changing any workflows.

6 Million Active Users: Scale Economics

OpenAI confirmed Codex has reached 6 million active users. This number matters for cost economics in two ways. First, scale allows OpenAI to amortize infrastructure costs across more users, creating room for price reductions or generous free tiers. Second, 6M users generating coding data creates a massive feedback loop for model improvement — better models need fewer tokens per task, which further reduces costs.

For context, GitHub Copilot reported around 1.8 million paid subscribers in early 2025. Codex's 6M figure (which includes free-tier users) suggests the market for AI coding assistance has expanded dramatically, and the subscription-vs-API pricing battle is intensifying.

What This Means for Your AI Coding Budget

The rate limit removal creates a temporary window where the effective cost per coding task drops significantly for subscription users. Here is how the math works across plan tiers:

• Plus ($20/month) — previously limited to roughly 40-60 Codex tasks per 5-hour window; now unlimited, making the per-task cost approach zero for heavy users
• Pro ($200/month) — already had higher limits, but removal eliminates the need to batch work or maintain API fallback budgets
• Business ($25/user/month) — teams can now run agent workflows continuously without hitting collective caps

The GPT-5.6 Family Cost Comparison

With the Sol efficiency updates, here is where the GPT-5.6 family sits relative to competitors for coding workloads:

• GPT-5.6 Sol — $5/$30, optimized for speed and coding, now 15-25% more token-efficient
• GPT-5.6 Terra — $2.50/$15, balanced reasoning and coding at lower cost
• GPT-5.6 Luna — $1/$6, lightweight tasks and high-volume batch processing
• Claude Fable 5 — $10/$50, premium reasoning with extended free window
• Claude Opus 4.8 — $5/$25, strong coding at moderate pricing

The pattern emerging is clear: providers are competing on effective cost per completed task, not just raw token price. Sol's efficiency improvements mean its $5/$30 pricing delivers more value per dollar spent than the headline number suggests.

Should You Switch During the Promotional Window?

If you are currently paying for API access to GPT-5.6 Sol for coding tasks, the unlimited Codex window represents a significant arbitrage opportunity. A Plus subscription at $20/month with unlimited Codex usage is dramatically cheaper than direct API usage that could easily exceed $200/month for active developers.

However, plan for the limits to return. OpenAI has described this as temporary, and the historical pattern suggests they will reimpose stricter limits once the promotional period ends — likely with higher caps than before but not unlimited. Budget for a return to rate-limited usage within 30-60 days and maintain your API fallback infrastructure.