Topic hub
Prompt Caching and Agent Costs
Briefs and tips that cover prompt caching, token pricing, rate limits and the cost of running AI coding agents, newest first.
Daily brief for 23 September 202623 Sept 2026
This is a price-and-routing day, not a capability-revolution day. Three labs priced long-running coding work into different buckets in about 36 hours, so teams that still point every Cursor, Claude Code, Codex, and CI agent at last month’s flagship are overpaying for bulk work…
Copilot Agent Permissions Meet Bulk Autofix and Effort Caps10 Sept 2026
GitHub Copilot can now pin enterprise deny/ask/allow rules for agent shell, file, and network operations. Code Quality agentic autofix lets you Assign to Copilot up to 25 backlog findings in one shot. Claude Code v2.1.267 adds `maxEffortLevel`, a prompt-snapshot toggle, and a large prompt-cache and resume fix pack. Gemini CLI v0.59.0 blocks …
Claude Corner: Set Prompt-Cache Breakpoints (512-Token Min)7 Sept 2026
Tip-only: place Anthropic prompt-cache breakpoints on the static tools and system prefix; clear the 512-token floor on Opus 5-class models.
GPT-6 Astra and Gemini Flash Force Cascade Routing as NVIDIA Buys Hugging Face5 Sept 2026
The last forty-eight hours compressed a year of stack decisions into one developer week. OpenAI began a phased rollout of GPT-6 Astra, a flagship it positions for software engineering, computer use, and cybersecurity, with API pricing of $10 per million input tokens and $50 per…
Daily brief for 3 September 20263 Sept 2026
Google ships Gemini 3.8 Flash as a coding workhorse at introductory Flash pricing through year-end. Claude Code 2.1.259 lets orgs push shared MCP servers and fail closed on unattended CI. LangChain 1.4 adds a first-party MCP adapter so servers become agent tools without custom glue. Copilot Business and Enterprise can lock a …
Claude Fable 5.1 and Gemini Flash slash agent costs as Astra stays gated2 Sept 2026
The last 24 hours were a billing and control event, not a quiet one. Anthropic put Claude Fable 5.1 on the general Claude Platform and the major clouds, pairing a new coding and research frontier with cheaper cache reads and a customer-owned review path still rolling out this…
Grok 4.6, Cursor, Claude, and Copilot close last gaps to production agents20 Aug 2026
Builders spent yesterday wiring long-horizon agents into the same clouds and IDEs they already operate. SpaceXAI’s Grok 4.6 reached Bedrock one week after launch, giving AWS shops a 500K-token coding and agentic model they can invoke with existing IAM, logging, and Cost Explorer…
LangSmith 82 Percent Cost Cut and Ornith-1.5 Self-Play Rival Frontier Agents19 Aug 2026
The agent engineering stack just got materially cheaper, more autonomous, and more native to the IDE. Production teams no longer have to sample a sliver of traces or pay frontier rates to know whether users are hitting silent failures. At the same time an emerging lab showed that…
Open Kimi K3 NotebookLM and19 Jul 2026
Today’s releases underscore a clear shift: frontier-scale models are becoming accessible for local customization and agentic coding at developer-friendly costs. Moonshot’s Kimi K3 stands out as the largest open-weight system yet, directly targeting long-running coding agents and…
Grok 4.5 GPT-5.6 Claude Cowork12 Jul 2026
## At a glance – xAI released Grok 4.5 this week, emphasizing strong coding performance and 4.2x token efficiency at competitive pricing. – OpenAI launched GPT-5.6 variants (Sol, Terra, Luna) on July 9 alongside ChatGPT Work for agentic productivity. – Anthropic responded with Claude Cowork for web and mobile on the same …
Sonnet 5 Hits 57 Percent on6 Jul 2026
## At a glance – Claude Sonnet 5 integration in Cursor delivers 57% on CursorBench, up from 49% with Sonnet 4.6, at introductory $2/$10 per million tokens pricing. – xAI launched the Grok Voice Agent Builder on July 1, a no-code platform for production voice agents with telephony, tools, and voice cloning. …
LangChain GitHub Updates Target4 Jul 2026
## At a glance – LangChain published practical guidance on July 2 for controlling exploding costs in production coding agents. – GitHub improved Copilot usage metrics reporting on July 2–3, giving teams better visibility into AI credit consumption. – Anthropic’s Fable 5 model returned to global availability on July 1 after regulatory …
Claude Sonnet 5 Memora MongoDB1 Jul 2026
## At a glance – Anthropic released Claude Sonnet 5 on June 30 as its most agentic Sonnet model, with major gains in coding, tool use, and reasoning at lower cost. – MongoDB added native reranking ($rerank) to Atlas in public preview, boosting RAG retrieval accuracy by ~24% on average without extra …
OpenAI Previews GPT-5.6 Family29 Jun 2026
## At a glance – OpenAI previewed the GPT-5.6 family (Sol flagship, Terra balanced, Luna fast/cheap) on June 26 with strong gains on agentic coding and cyber benchmarks, but initial access is government-gated to ~20 vetted partners. – New tiered pricing strategy positions Terra at roughly half the cost of GPT-5.5 while …
xAI Grok 4.3 and OpenAI Agents2 May 2026
## At a glance xAI ships Grok 4.3 to the public API with 1M-token context, always-on reasoning, and ~40-60% lower pricing for agentic workloads. Grok Imagine Agent Mode beta launches on web, turning image/video generation into autonomous, canvas-based iteration. OpenAI Agents SDK adds native sandbox execution, file tools, and harness support for …