A new guide on optimizing AI agent cost, speed, and quality recommends routing individual agent steps among GPT-4o-mini, GPT-4o, and GPT-5.6 Luna based on task complexity, tool requirements, confidenc…
A comparison published by Entelligence benchmarked GPT-5.6 Luna and GPT-6 Astra on identical pull requests, finding Luna cost $0.0041 per review versus Astra's $0.113 but caught 69 verified bugs at 74…
GPT-5.6 Luna found 69 verified bugs across 50 public benchmark pull requests at a total cost of $0.20, while GPT-6 Astra found 92 at $5.66, according to an evaluation published by the AI-Code-Review-E…
Databricks released Adaptive Instructed-Retriever, a retrieval model that dynamically adjusts how many sequential search steps it performs based on query difficulty, letting developers cap the maximum…
ArchBench launched as a desktop AI assistant that controls Blender through chat, letting users describe a modeling change, review the proposed action and affected objects, then apply it in the native …
AWS announced August 2026 updates to Amazon Bedrock, AgentCore, and the open-source Strands Agent Harness SDK, including million-token context windows for GPT-5.6 Sol, Terra, and Luna models, web sear…
OpenAI released GPT-6 Astra on September 3, introducing four breaking API changes that will cause existing integrations to fail if not addressed before migration. The changes include the removal of sa…
Simon Willison's comparison grid of AI-generated pelicans shows OpenAI's GPT-6 Astra produces significantly better images than GPT-5.6 Sol, Terra, and Luna across all reasoning levels, with Astra low …
OpenAI's GPT-6 Astra caught approximately 4% more labeled bugs through actionable findings than GPT-5.6 Sol and 22% more than Opus 5 in CodeRabbit's evaluation, with gains reaching 20% over Sol and 33…
OpenAI has released GPT Astra, a new model family including GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna, designed for complex reasoning, balanced intelligence and cost, and cost-sensitive high-volume…
In an experiment with three small datasets, OpenAI's GPT-5.6 Terra and GPT-5.6 Luna models made repeated data-analysis errors, including approving a backwards conclusion and inventing a correction tha…
Amazon Bedrock released GPT-5.6 Luna (Global) on 2026-07-09, a model with 1,050k context and 128k max output, supporting text, image, and PDF inputs. The model is available under the identifier global…
Amazon Web Services announced that Australian teams can now access OpenAI's GPT-5.6 Sol, Terra, and Luna models through Amazon Bedrock with global cross-Region inference from the Asia Pacific (Sydney)…
Researchers introduced CordisBench, a 1,200-question benchmark testing whether language models can reason about component lifecycles in dynamic agent harnesses, finding that models handle small system…
Anthropic's Claude Fable 5.1 has topped the Artificial Analysis Intelligence Index with a score of 66, the highest ever recorded, but at $3.69 per task it is the most expensive model tracked, costing …
Inception released Mercury 2.5, a diffusion large language model (dLLM) that generates tokens in parallel, achieving 1,107 tokens/sec on standard GPUs and a 10+ point intelligence jump over Mercury 2,…
Anthropic released Claude Fable 5.1, now available in Kilo, with cache read prices cut 75% to $0.25 per million tokens and benchmark gains on Terminal-Bench 4.0 (42% to 55.8%) and AutomationBench (17.…
Wren AI has switched its default model to GPT-5.6 Luna, cutting cost per task by 16× from $0.58 to $0.036 and reducing model time by 37% from 161.9s to 102.5s, while maintaining 95% accuracy on evalua…
Inception released Mercury 2.5 Preview on August 31, 2026, claiming 1,107 tokens per second on standard GPUs, a 10-point intelligence jump over Mercury 2, and quality comparable to GPT-5.6 Luna at low…
Artificial Analysis released a comparison of 13 frontier AI models, ranking them by fidelity, cost, and speed, with GPT-5.6 Sol and DeepSeek V4 Pro 0813 achieving the highest fidelity (99-100), while …