AI Briefing#token-usage
The latest AI and developer news about #token-usage, with the original source and a short summary.
The latest AI and developer news about #token-usage, with the original source and a short summary.
An effort_cost_index, representing relative token usage estimates for Effort levels by model, was discovered in the Claude Code binary.
Analysis reveals that actual reasoning token usage in Claude models is significantly lower than benchmark levels, creating an 'Inference Gap' and prompting calls for transparency and regulatory discussions.
Claude Code's Artifact feature consumes approximately 6,000 tokens per session and can be disabled in the settings file.
OpenAI's GPT-5.6 discount policy increased token usage for the Terra and Luna models by 5.6x and 13.8x respectively, eroding competitors' market share.
SessionTrim reduces storage space by removing tool outputs and system-only data from Claude Code and Codex sessions.
Uber and Walmart have introduced usage limits to manage rapidly rising AI costs.
GPT-5.6 Sol's per-session token usage increased 2.25x compared to GPT-5.5.
CodeBurn, an open-source dashboard that locally analyzes usage and costs across 36 AI coding tools, has been launched.
According to Vercel data, Open-Weight models accounted for 29% of AI Gateway token usage while showing high efficiency.
A comparative analysis of token usage and cache efficiency between Claude Code and OpenCode found that Claude Code's overhead was far higher.
It has been revealed that the actual number of tokens consumed to complete a task, rather than the per-token price of a model, is the key variable determining cost.
Beyond raw performance, 'intelligence per dollar'—a measure of AI model cost efficiency—is emerging as a new industry standard.
Uber's COO pointed out the cost-efficiency problem arising from maximizing LLM token usage.
The creator of OpenClaw ran 100 coding agents and spent $1.3 million in OpenAI API costs in a single month.
In Claude Code, Opus 4.7 serializes independent file reads, causing a large increase in token usage.
In Claude Code, Opus 4.5, 4.6, 4.7, and Sonnet 4.6 were compared by cost and tokens across effort levels.
In the same email classification agent, Opus 4.7's token usage and cost increased.
Agents, skills, and MCP piled up, consuming 30,000 tokens before the session even started.
A plugin that tracks and manages Claude Code Skill usage and token consumption.
A cache_read calculation error in the Pro Max 5x plan causes quota to be consumed rapidly.
Meta's internal token leaderboard revealed 60 trillion tokens used in 30 days and the flaws in the metric.