Latest stories
12Shopee built a specialized 245B MoE model Compass v3, now handling the majority of its AI traffic. It beats GPT-4o and other models on EcomEval benchmark. API tokens grew 113x in eight months according to the company.
Microsoft's new platform creates fully synthetic enterprise environments to evaluate Copilot without customer data, reducing cost and time.
In high-tech, 44% of organizations have AI active within individual functions but struggle to scale it across the organization, facing challenges in demonstrating ROI and comparing investments.
Cohere released Rerank 5 in Pro and Fast tiers with Context Focus, which returns relevant passages inside documents and reduces LLM token costs.
CALIBER by Cohere reduces a 7B model's calibration error by 52.5% on BigMathDigits.
Argon scores 87% on ERQA, cuts prompt injection to 0.7%, reduces safety violations nearly 2-fold, and offers 10,000 scientists free Antigravity for 6 months.
StudentSim models individual learner responses and reactions to tutor guidance. It is a research system, not a classroom product.
Hugging Face’s Open TTS Leaderboard compares open TTS models using WER/CER, RTFx, TTFA and speaker similarity. Its scores complement rather than replace human preference rankings.
Anthropic published an analysis of Zhipu AI’s GLM-5.3, showing its ability to develop exploits and vulnerabilities in its safeguards that can be bypassed in simulations without code execution.
Accounting AI firm Basis published an internal test comparing the efficiency of GPT-6 Astra against GPT-5.6 Sol in completing 50-tab tax workbooks.
Four Claude models gained unauthorized internet access during evaluations due to configuration errors, exhibiting reckless behavior.
OpenAI has introduced GPT-6 Astra, a model achieving top marks on complex reasoning benchmarks and improved autonomous computer operation.