Week 32, 2026

Seedance 2.5 is ByteDance's video generation model, launched 31 July 2026 inside Jimeng AI and Doubao Pro. It generates clips up to 30 seconds at 4K in a single run, producing video and audio together in one pass rather than dubbing audio afterwards, and supports multi-turn extension for longer sequences. Its distinguishing feature is an input system accepting up to 50 multimodal references at once, images, text descriptions, style frames, character references and scene direction, used to steer character and style consistency across a shot. It is served through Volcano Engine Ark and BytePlus, which publish model IDs and per-token pricing, though rollout has been staged rather than open to all developers at once.

Claude Opus 5 is Anthropic's flagship model, released 24 July 2026 with a 1M-token context window and five selectable effort levels (low, medium, high, xhigh, max). Effort is the main control: output token spend runs roughly 8x from low to max, and Artificial Analysis measures a 407-Elo spread in task quality across that range, so the same model behaves like several different price and capability tiers. API pricing is $5 per million input tokens and $25 per million output, with cache writes at $6.25 and cache hits at $0.50. It leads the Intelligence Index at 60.7 and tops the Coding Agent Index, scoring 89 percent on Terminal-Bench v2.1 at max effort and 53 percent on Humanity's Last Exam.

OpenArt is a hosted generative image platform with a node-based workflow builder alongside a conventional prompt interface. Workflows can be built from templates or from scratch, run on hosted compute, and published for other people to reuse.

Weavy is a browser-based node canvas for chaining hosted generative models into a single pipeline, mixing image, video and editing steps from different providers in one graph rather than moving files between tools.

Invoke is an open-source generative image platform combining a unified canvas with inpainting, outpainting and layer control, plus a node editor for building repeatable workflows. It can be self-hosted or used through a commercial hosted tier aimed at studios.

Muse Spark 1.1 is Meta Superintelligence Labs' multimodal reasoning model, released 9 July 2026. It has a 1M-token context window, accepts text, images, video and PDFs, and is built for agentic work: tool use, computer use, coding, and multi-agent orchestration. It is free to use in the Meta AI app and at meta.ai in Thinking mode, and available to developers through the new Meta Model API at $1.25 per million input tokens and $4.25 per million output, with $20 in starting credits. Unlike the Llama family it succeeds in practice, Muse Spark is closed-weight.

Nemotron 3 Ultra is the largest member of NVIDIA's Nemotron 3 family, released 4 June 2026 at Computex. It is a 550B-parameter hybrid latent mixture of experts with roughly 55B parameters active per token, combining Mamba and Transformer blocks, and trained in NVIDIA's 4-bit NVFP4 format on Blackwell hardware. It ships under the NVIDIA Open Model License, which permits commercial use, and NVIDIA published training data, reinforcement learning environments and post-training recipes alongside the weights rather than the weights alone. Weights are on Hugging Face as nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B in both BF16 and NVFP4, and it is served through OpenRouter, Together AI, Baseten, DeepInfra, Fireworks and NVIDIA NIM.

FLUX 3 is Black Forest Labs' multimodal foundation model, announced 23 July 2026. Unlike the FLUX.1 and FLUX.2 image models before it, FLUX 3 learns jointly across images, video and audio in a single unified architecture: it generates video with native synchronised audio, edits images, renders readable text, and, via a FLUX-mimic variant, predicts robot actions, all from the same weights. Video generation runs up to 20 seconds. At launch, video is available through a gated early-access programme, with image generation stated to follow and an open-weight FLUX 3 Dev backbone planned later.

GLM-5.2 is Zhipu AI's (Z.ai) flagship open-weight model, released 13 June 2026 under an MIT licence with weights published on Hugging Face at zai-org/GLM-5.2. It is a 753B-parameter mixture-of-experts model activating roughly 40B parameters per token, with a 1M-token context window and 128K maximum output. The headline architectural change is IndexShare, which reuses the same indexer across every four sparse attention layers; Z.ai reports this cuts per-token compute by about 2.9x at full 1M context. It targets long-horizon agentic coding, multi-file refactors, terminal work, and tasks that run for many steps rather than single completions.

Inkling is the first open-weights model from Thinking Machines Lab, released 15 July 2026 under Apache 2.0. It is a 975B-parameter mixture of experts with 41B active per token, trained from scratch on 45 trillion tokens of text, images, audio and video, with a context window up to 1M tokens. The architecture uses 256 routed experts plus 2 shared experts per layer with 6 routed experts active per token, a sigmoid router, and interleaved sliding-window and global attention at a 5:1 ratio. It accepts text, image and audio input natively and returns text. Weights are on Hugging Face in both the original format and an NVFP4 checkpoint for Blackwell hardware. A distilled Inkling-Small followed on 31 July.