Related News
Nov 10, 2026
Google's MKA benchmark tests AI agents on 200 private professional tasks; a separate set of 10 public tasks is available for inspection
Google DeepMind's new Multimodal Knowledge Agent benchmark evaluates AI agents on 200 private tasks requiring visual analysis of complex professional files like architectural blueprints and engineering schematics. While the benchmark's headline scores are based on these private evaluations, only 10 sample tasks are publicly available for inspection. The benchmark uses weighted criteria where harder 'Climb' criteria measure advanced performance, and costs reported by Artificial Analysis exclude judge calls, limiting direct workplace cost comparisons.
Nov 8, 2026
Inkling 2 Small: Hosted on Tinker Now, Open Weights Planned Next Week
Thinking Machines Lab's Inkling 2 Small, a 594B-parameter MoE model with 17B active per token, is available today via Tinker's hosted API with closed weights. The company intends to release the full weights under an Apache 2.0 license next week after safety evaluations, relative to its November 5 announcement. The model supports text and image inputs (no audio), offers 64K and 256K context variants, and its per-token pricing on Tinker is operation-specific: prefill costs $0.30 per million tokens ($0.03 cached), sampling costs $1.00, and training costs $1.50 for the 64K context version. This article clarifies the current deployment routes, exact context limits, input modalities, and cost structure to help developers choose between immediate hosted access and future self-hosting.
Oct 29, 2026
Microsoft’s Copilot harness comes to Foundry for managed agent execution
Microsoft’s October 28 announcement separates the model from its instructions, context, execution and controls, with managed Foundry and downloadable SDK routes.