Northflank warns all customers to rotate secrets after AI keys were abused in October breach
The affected-customer count and entry point remain unconfirmed. Revoking keys at the issuing provider matters more than simply changing a platform variable.
Search
Search results for “edge inference economics”
Find published research by company, product, platform or technology.
Showing 61–80 of 241 dossiers
Kimi K3.1 Pro reaches frontier-tier agent benchmarks at $2/$10 — weights are promised, not shipped
Kimi K3.1 Pro ships via API at $2 input/$10 output per million tokens, with a 2–3x faster Highspeed tier. Independent Artificial Analysis testing scores 55 on its Intelligence Index and 67% on Terminal-Bench 4.0. Open weights remain promised, not released.
LaunchDarkly AgentControl now versions and deploys agent skills without application releases
Agent skills join prompts and models as managed runtime configurations; approvals and pinned versions make changing agent behaviour auditable rather than an untracked edit to skill files.
Artificial IntelligenceCloud & InfrastructureAI ModelsAI AgentsCloud PlatformsInference & APIsActive dossier
Grok 4.6 reaches AWS, Google and Microsoft clouds within two weeks
Grok 4.6’s distribution expanded unusually quickly after launch. Builders can now evaluate and deploy the model through AWS, Google and Microsoft enterprise AI platforms while keeping each cloud’s existing governance, logging and regional-control layer.
Artificial IntelligenceWeb DevelopmentCloud & InfrastructureAI AgentsDeployment & DevOpsDeveloper ToolsObservabilityActive dossier
Vercel Agent is bringing production investigations and approved actions into Slack
Vercel Agent now works in Slack as well as the Vercel dashboard, combining logs, metrics, deployments and repository context with team conversation before proposing approved actions such as pull requests, rollbacks, configuration changes and cache purges.
Cloudflare opens Quicksilver-backed KV Instant — fast global reads with $100-per-MB storage
Workers KV Instant is built for hot-path flags, not general storage: Cloudflare quotes 1.62ms p99 reads, $0.20 per million reads, $0.10 per write and $100 per MB each month. Private beta limits are strict.
Cloudflare makes post-quantum origin TLS automatic — and cuts p90 handshake latency by more than 150ms
The useful shift is automation at the CDN-to-origin boundary: operators no longer need to manually force post-quantum key exchange, while Cloudflare says its measured HelloRetryRequest rate fell from about 52% to 3.7% across the scanned cohort.
Cloudflare Workers now enables Node.js compatibility by default for new compatibility dates
Cloudflare has inverted a long-standing Workers assumption: Node.js compatibility is now on by default for current compatibility dates. That reduces setup friction for many npm packages, but it also means developers need to understand compatibility-date boundaries, partial APIs and explicit opt-out flags.
Xiaomi opens MiMo-V2.6 weights and the RL machinery behind its frontier-scale training run
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.
Mistral Large 4 weights are downloadable — but hosted model services need a separate licence
The trillion-parameter Mistral Large 4 is no longer API-only; builders can self-host, but must assess licence and hardware limits.
Cloudflare tests post-quantum regional key fencing — but CDN protection is not ready yet
PQ GeoKey Manager uses hybrid cryptography, regional root keys and fail-closed key unwrapping. General availability is planned for Q1 2027; CDN integration and independent assurance remain outstanding.
Google Play’s August 31 target-API deadline can block updates and hide older apps from new users
Google Play’s 2026 target-API cutoff has two separate consequences: most new submissions and updates need API 36, while existing apps below API 35 can lose distribution to new users on newer Android devices. Developers who need more time can request an extension to November 1.
GPT-6 Sol and Luna cut OpenAI’s mid-tier API prices in half
The important change is economic rather than another flagship benchmark win. OpenAI is making capable agent and coding workloads materially cheaper, with Luna approaching older Sol-class results at a tiny fraction of the task cost and GPT-6 prompt caching discounting reused input by up to 90%.
OpenAI frontier models reach Google Cloud — with Marketplace billing against cloud commitments
Google Cloud customers gain a new route to OpenAI models with unified governance and cloud-commitment billing, but broad availability and some privacy features are still forthcoming.
Cloudflare opens Logpush to free plans — with 25GB allowances and pay-as-you-go export
Cloudflare's logs are no longer an Enterprise-only export capability. Small sites can send 25GB a month to internal destinations and another 25GB externally before overage charges, but destination costs and separate Workers/OTel meters still matter.
DeepSeek’s current API docs contradict the announced V4 Pro-to-Flash reroute
The live DeepSeek changelog and rate card still show distinct V4 Pro service after the previously announced September 14 reroute. That changes cost and model-selection assumptions.
GLM-5.3-Flash turns the anonymous Ox Alpha trial into an open-weight multimodal coding model
GLM-5.3-Flash combines open weights, multimodal coding/agent capability and an 18B-active sparse architecture with a large anonymous pre-launch trial. Z.ai has already issued a chat-template correction for early downloads, showing that day-one self-hosted deployments need artifact-level validation as well as model benchmarking.
K2 Horizon opens more than model weights — including training data, code and intermediate checkpoints
K2 Horizon is notable less for another benchmark claim than for reproducibility: IFM is publishing model weights, architecture, training code, data or construction recipes, evaluation resources and intermediate training material instead of stopping at a final checkpoint.
Claude Opus 5.5 brings Fable-class performance into a cheaper frontier tier
The interesting change is economic as much as benchmark-driven. Anthropic is compressing capability that previously justified its larger Fable tier into Opus pricing, while cutting Opus list prices and expanding immediate availability across the major clouds.
Abacus.AI’s Smaug Agentic fine-tune targets the failure tail in long-running coding agents
The useful part of Smaug Agentic is not another frontier-style benchmark claim. Abacus.AI is publishing a drop-in Kimi K3 derivative that targets a specific production failure mode in coding agents: long runs that burn the reasoning budget without converging. The weights and model card are public, but the training data is not disclosed and the benchmark gains remain vendor-produced.