Find published research by company, product, platform or technology.
Showing 361–373 of 373 dossiers
MiMo-V2.6 is more useful than another benchmark launch because builders get both capable multimodal weights and a rare view into the reinforcement-learning machinery that produced them: code, environments, run costs and even failure notes from the training cluster.
The architecture matters as much as the voice quality: developers can replace a chained speech-to-text → LLM → text-to-speech loop with one full-duplex conversational model while keeping their own choice of backend reasoning model, tools and agent harness.
The technical migration is unchanged, but Reddit's documentation has proven unusually volatile. Integrations should now treat September 21 as the current objective/form readiness date and December 8 as the current published cutoff for HOUR reports longer than seven days—while continuing to recheck the live docs before each deadline.
Muse Voice Transcribe gives voice-app builders one streaming model for transcription, speaker separation and turn detection instead of stitching those stages together. Its low published price is notable, but Meta’s benchmark claims still need workload-specific validation.
Cloud Run instances sit between autoscaling serverless services and a small VM. They run one individually addressable container continuously, can be stopped and restarted, and use shared CPU economics; Google’s launch example prices 1 vCPU plus 1 GiB running for 30 days at $5.70.
The Imagen 4 shutdown is now effective, not merely scheduled. Builders still calling the old model IDs need to migrate to current Gemini image generation, where model names and interaction patterns differ enough to warrant explicit compatibility testing.
Replit’s August 2026 Cloud pricing changes materially lower several production costs: autoscale compute falls from $3.20 to $0.60 per million compute units and database storage from $1.50 to $0.35 per GiB-month. The details matter because not every SKU moved down.
Hy4 preview is a very large sparse model with public full and FP8 weights, native speculative decoding and a 1M-token context path. Its open release makes Tencent’s claims testable, while the 1.56TB full checkpoint keeps self-hosting firmly in server-scale territory.
CS-4 combines three WSE-3 Turbo wafers with Cerebras’ Nexus rack design. The practical shift is architectural: compute, power and I/O become modular, while Cerebras now says the same platform is intended to support CS-5 in 2027 and a 3D-memory CS-6 generation after that.
Groq 3 LPX is moving from architecture announcement to manufactured infrastructure. Artificial Analysis measured about 3,400 output tokens/s at both 10K and 100K context on an NVIDIA-hosted private endpoint, but the single-concurrency benchmark does not yet establish public-cloud price, multi-tenant throughput or end-to-end agent speed.
AWS has added a `REFERENCE` mode for Lambda deployment packages. It eliminates duplicate managed copies, raises the default managed-storage quota to 300GB, and gives teams direct control over encryption, lifecycle and audit policy—but a deleted or inaccessible source object can now make a function inactive.
Cloud Run sandboxes now cover all resource types. The August 5 expansion matters for builders whose agents or automation run in batch jobs or continuously pulling workers rather than HTTP services, while the feature remains pre-GA and shares CPU and memory with the host container.
Fin’s new Evals and Releases features let teams test agent changes against simulated conversations before publishing, bundle configuration into a release, ramp traffic or A/B test it, and feed failures from live Monitors back into the next iteration.