Continuous Intelligence Releases ci-proto, an Open Time-Series Forecast Model
Continuous Intelligence released ci-proto on 12 November 2026, a 167M open-weight time-series model. The lab and a public board differ on its fev-bench rank.

Quick coverage of new AI model releases. Recent posts cover Google's Gemini Omni 1.1 Flash with scene extension and keyframe control, Meta's Muse Code agent and Muse Spark 1.2, and Z.ai opening GLM-5.3 API access with prompt caching.
Continuous Intelligence released ci-proto on 12 November 2026, a 167M open-weight time-series model. The lab and a public board differ on its fev-bench rank.

Google opened Gemini 4 Carbon to paid API users, Gemini for business and Google AI Ultra. It costs $2/$10 per 1M tokens until Nov 30, then $4/$20.

MiniMax released open weights for H3-Image, a 16-megapixel text-to-image and editing model. Its license excludes the EU, UK, South Korea and US.

NVIDIA posted four 16B Cosmos 3 Nano open checkpoints for interactive world simulation on Hugging Face under OpenMDW-1.1, for driving and robot sims.

Ant Group launched Ling-3.2-flash on Nov 5, 2026: 560B total, 25B active, 1M context, image and video input, free for two weeks, no open weights yet.

Baidu released ERNIE-5.5-Flash-Base on Nov 5, 2026: a 280B MoE with 14B active per token, 1M context and a custom license that excludes the EU and UK.

Klein-TTS X1.5 and VoxRT's LinnetTTS are small text-to-speech models for devices. Sizes, languages, speeds and licenses from the vendors' own pages.

Meituan released LongCat-2.5, a 1.6T-parameter model it says was trained entirely on Chinese accelerators. Its price, benchmark claims and open questions.

Soniox STT v6 lets a voice agent send its own words to the speech model so replies are recognized better and the agent's echo is filtered. Price and limits.

Edge0's Audio8-TTS-Realistic-Air is a 0.6B Apache 2.0 text-to-speech model with zero-shot voice cloning in 23 languages. What the card and paper report.

Tencent released Hy4 on November 4, 2026 with image input and computer use, the Preview's price, and open weights promised in the coming days.

Moonshot posted Kimi K3.1 Pro weights on Hugging Face: 2.8T parameters, 104B active, 1M context, about 1.56 TB of files and a $20M revenue term for API hosts.

OpenBMB's Apache 2.0 MiniCPM-OCR 1.0 reads scans, tables and formulas at 1.2 pages per second on one GPU, and a 2B Jev model scores OCR output for review.

Canva Brush, released October 29, 2026, is a vision-language model that edits Canva designs from a prompt. How it works, Canva's claims and how to try it.

Baseten's MIT-licensed joey-1 returns calibrated yes/no, choice and score probabilities in one pass, at $0.01 per million input tokens. Claims and limits.

Pika H3 Flex is a MiniMax H3-based video model with open weights, up to 12 references, five audio stems and hosted 4K upscaling. Specs, prices and limits.

ServiceNow's MIT-licensed Apriel Voice 27B Preview takes speech in and returns text and tool calls. Its benchmarks, hardware needs and limits, explained.

IBM's Granite Switch 4.1.1 adds multilingual speech input and new Decision adapter weights, while the base model stays untouched.

Ideogram Vector 1.0 turns a prompt or an existing image into an editable SVG. What it does, why vector output matters, and API prices of $0.10 to $0.32.

DiffuSpace's Clover-35B-A3B-Preview is an Apache 2.0 diffusion language model built from Qwen3.6-35B-A3B. How it works, its claimed scores and how to run it.
