OpenAI Codex July 20: CLI 0.144.6 Refreshes GPT-5.6 Sol Instructions With Corrected 272,000-Token Context, 8 Million Active Users, Usage Limits Reset Again

OpenAI Codex hits 8 million active users, usage limits reset for all, 5-hour cap stays off for GPT-5.6 Sol

• On July 14, Codex lead Thibault Sottiaux announced 8 million active users across Codex and ChatGPT Work, reset usage limits for all users once again, and confirmed the 5-hour rate limit remains off so developers can push GPT-5.6 Sol on ambitious tasks.

Editor’s read: The repeated resets are a growth play - use the headroom while it lasts, but do not architect workflows around limits staying off.

Codex CLI 0.144.6 refreshes GPT-5.6 Sol, Terra and Luna instructions and corrects context windows to 272,000 tokens

• On July 18, rust-v0.144.6 shipped refreshed bundled instructions for GPT-5.6 Sol, Terra and Luna and corrected their context windows to 272,000 tokens, a metadata fix that changes how much context the CLI will actually pack into a session.

Codex CLI 0.144.5 improves dangerous-command detection and explains why commands are denied

• On July 16, rust-v0.144.5 improved dangerous-command detection.
• including more forced rm forms.
• and now gives clearer rejection reasons when a command is denied, so you can tell a safety block from a bug.

OpenAI Codex Micro: a $230 light-up keyboard for managing Codex agents

• On July 15, OpenAI released the Codex Micro, a $230 limited-run keyboard co-designed with Work Louder: light-up Agent Keys show agent status, Command Keys map to frequent Codex actions, a joystick launches workflows, and a dial adjusts agent reasoning levels, all controllable from the ChatGPT desktop app.
• The launch landed while OpenAI contests Apple's hardware trade-secret lawsuit.

Editor’s read: A novelty, but a telling one: OpenAI is betting developers will manage fleets of agents, not single sessions.

Claude Code vs Codex vs Cursor vs Grok Build: the July 2026 state of play for AI coding agents

• On July 19, claude-world.com published a market snapshot arguing there is no universal winner: Codex spans desktop, CLI, IDE, cloud and SDK surfaces with configurable subagent nesting and a stable codex exec command for non-interactive automation, and the right choice depends on workload, failure tolerance and scale.

Claude Code, Codex and Cursor usage limits all reset on July 16

• On July 16, three major AI coding platforms refreshed quotas the same day: OpenAI refreshed Codex allocations, Anthropic reset Claude Code's rolling windows, and Cursor doubled included usage for Grok 4.5 and Composer 2.5.
• a coordinated retention war among frontier labs.

Kimi K3: Moonshot AI's 2.8 trillion parameter open-source model tops a major coding benchmark

• On July 18, coverage of Moonshot AI's Kimi K3 highlighted the largest open-weight model yet.
• 2.8 trillion parameters, one-million-token context.
• placed behind GPT-5.6 Sol and Claude Fable 5 overall but with a 76% win rate on the Frontend Code Arena; open weights are promised by July 27, 2026.

Gemini 3.5 Pro misses its third launch deadline as Google eyes a stopgap Gemini 3.6 Flash

• On July 15, reports confirmed Gemini 3.5 Pro missed its July 17 general-availability target.
• the third slip since June.
• leaving GPT-5.6, Claude and Grok as the current frontier options for agentic coding while Google regroups.