GPT-5.6 Luna

Headlines · 6

1. 3d
   GPT-6 Luna system card addendum updates MentalHealthBench results

   An October 29, 2026 addendum to the GPT-6 Luna System Card reports MentalHealthBench results for the October 29 version. The model scored 48.68 overall, 46.39 on non-acute cases, 53.92 on high-acuity cases and 49.64 on emergent cases. Scores use the task-clipped metric on a 0–100 scale, where higher is better, and are based on four responses per task. The supplied report does not explain what changed between the October 7 and October 29 versions.

   OpenAI Deployment Safety · 🔥 0
2. 3d
   Open-weight models dominate tokens, not OpenRouter spending

   OpenRouter data cited by Goldman Sachs’ Bobby Molavi shows open-weight models handled 72% of tokens in the second week of September but received only 27% of platform spending. The report also says token demand rose sharply while GPU rental prices increased and token prices fell, creating a disputed divide between infrastructure revenue and model monetization. It further cites roughly 69% of mid-September tokens as coming from Chinese AI laboratories and reports that Meta sought up to $95 billion from shareholders for AI capital expenditure.

   36氪 人工智能 · 🔥 0
3. 4d
   Retell AI runs 100M monthly calls on OpenAI models

   OpenAI reports that voice-AI platform Retell AI now serves more than 4,000 businesses, with monthly call volume growing from 30 million in early 2026 to over 100 million by September, and OpenAI models generating responses for 80% of those calls. Retell says customer resolution rates exceed 80%, citing Anker's reported 80.4% case resolution rate and a Net Promoter Score of 63, while Pine Park Health's scheduling agent reached a 55.7% booking rate. In its vCX benchmark dated October 13, 2026, GPT-6 Astra ranked first for instruction following at 89.7%, and Retell engineers use Codex for code research, feature building and debugging plus GPT-6 Astra's computer use to test front-end interfaces.

   OpenAI · 🔥 0
4. 5d
   AWS Explores Prompt Caching Patterns for AI Agents on Bedrock

   The AWS Machine Learning Blog explains how Amazon Bedrock prompt caching can reuse previously processed context, including instructions, tool definitions, and conversation history, instead of resending the full context on every agent step. The guide covers implicit and explicit cache breakpoints, stable prefixes, additional tools, cache eligibility, and write costs. In one six-turn workload, AWS reported about 62% lower input cost with implicit caching and about 61% with explicit rolling breakpoints on GPT-6 Astra, while cautioning that results vary by workflow.

   AWS Machine Learning Blog · 🔥 0
5. 6d
   OpenAI and Anthropic Cut LLM Prices by Up to 80%

   Max Woolf’s analysis charts how OpenAI and Anthropic reduced API prices for major models between June and September, including an 80% cut for GPT-5.6 Luna and lower prices for GPT-6.1 Sol and Claude Opus 5.5. The reports say competition from lower-cost open-weight models, including DeepSeek and Kimi offerings, created pricing pressure, while improvements in inference efficiency may also have enabled lower prices. Price comparisons in the analysis cover input, output and cache-read charges per million tokens.

   Hacker News · AI（100+ 分） · 🔥 0
6. Oct 20
   Amazon Bedrock brings OpenAI GPT models to Canada

   AWS says Amazon Bedrock cross-Region inference now makes GPT-5.6 Sol, Terra and Luna, GPT-6 Astra, Sol and Luna, and GPT-6.1 Sol available from Canada Central and Canada West (Calgary). Applications call the local Bedrock Runtime endpoint, while requests are routed to supported destination regions for inference; the models are accessible through the OpenAI Responses, Chat Completions and Bedrock Converse APIs. AWS says North American inference and data-residency controls are applied, while flagged content may be retained in the destination region for up to 30 days.

   AWS Machine Learning Blog · 🔥 0

Experience and discussion from the community

Share my GPT-5.6 Luna experienceAsk about GPT-5.6 Luna