This tracker follows every major AI model released in 2026. We include models from open-weight releases to proprietary APIs, covering language models, multimodal models, and specialized models. Each entry links to our full coverage article.
What counts as a "major" release: A model that changes the competitive landscape. This includes top-tier benchmarks, open-weights that rival proprietary models, or a new capability tier such as multimodal or agentic reasoning.
Model Release Table
| Model | Company | Type | Key Capability | Status |
|---|---|---|---|---|
| Claude Opus 5 | Anthropic | LLM | Safest model at half the price of Opus 4 | Released Aug 22 |
| DeepSeek V4 Flash Vision | DeepSeek | Multimodal | Vision plus reasoning, challenges Claude | Released Aug 22 |
| GPT-5.6 Sol | OpenAI | LLM | 20% price cut triggers AI price war | Released Aug 23 |
| Inherent Faraday | Inherent AI | LLM | Beats OpenAI and Anthropic at research | Released Aug 23 |
| Thomson Reuters Legal AI | Thomson Reuters | Domain-specific | Proprietary legal reasoning model | Released Aug 25 |
| Groq 3 LPX | NVIDIA/Groq | Inference chip | 3,400 tokens/sec for agentic AI | Released Aug 25 |
| Ox Alpha | Z.AI | LLM | Open-source weights coming | Confirmed Aug 26 |
| ChatGPT Ads India | OpenAI | Product | Free and Go tiers with ads | Released Aug 27 |
| Claudeforce | Anthropic/Salesforce | Enterprise | Claude-powered CRM model | Released Aug 27 |
| DeepSeek V4 | DeepSeek | Multimodal | Agent harness, multimodal reasoning | Released Aug 28 |
| Tencent Hy4 Preview | Tencent | LLM | 770B open-weight flagship | Preview Aug 28 |
Timeline
August 29, 2026: Tencent Hy4 Preview: 770B open-weight flagship
Tencent previewed Hy4, a 770 billion parameter open-weight model positioned as its flagship LLM. At 770B parameters, it enters the largest open-weight category, competing with Meta's Llama series and Mistral's largest releases. The preview suggests it targets research and enterprise use cases. Read the full article.
August 28, 2026: DeepSeek V4: Open weights, agent harness, multimodal
DeepSeek released V4, a 236B MoE model with open weights, agent harnessing capabilities, and full multimodal support. It challenges Claude Opus 5 across reasoning and vision tasks while providing open-weights access that the competition does not offer. Read the full article.
August 27, 2026: Claudeforce (Salesforce plus Anthropic)
Salesforce and Anthropic announced Claudeforce, a Claude-powered enterprise AI model built into Salesforce's CRM platform. It represents the deepening integration of LLMs into enterprise software workflows. Read the full article.
August 27, 2026: OpenAI announces ChatGPT ads in India
OpenAI launched a free tier of ChatGPT supported by ads, targeting the Indian market. The Go tier offers a lower-cost option for price-sensitive users. This is a product expansion rather than a new model. Read the full article.
August 26, 2026: Z.AI confirms Ox Alpha: open-source weights coming
Z.AI confirmed its Ox Alpha AI model and pledged to release open-source weights. This follows a growing trend of Chinese AI labs, including DeepSeek and Tencent, competing with US models through open-weight releases. Read the full article.
August 25, 2026: NVIDIA DLSS 4.5 and Groq 3 LPX
Two infrastructure announcements: NVIDIA shipped DLSS 4.5 Ray Reconstruction for 30-plus games, and released the Groq 3 LPX inference accelerator delivering 3,400 tokens per second for agentic AI workloads. Read the DLSS article. Read the Groq article.
August 25, 2026: Thomson Reuters launches proprietary legal AI model
Thomson Reuters launched its first proprietary legal AI model, Thomson, designed for legal research and document analysis. It is not open-weight and targets the enterprise legal market. Read the full article.
August 23, 2026: Inherent Faraday beats OpenAI and Anthropic at research
Inherent AI released Faraday, a model that reportedly outperformed both OpenAI and Anthropic's top models on research benchmarks. The claim is notable because it comes from an unknown company challenging the established duopoly. Read the full article.
August 23, 2026: OpenAI cuts GPT-5.6 Sol prices 20%
OpenAI cut GPT-5.6 Sol prices by 20%, triggering an AI price war. The move signals intensifying competition in the API model market and puts pressure on Anthropic, Google, and others. Read the full article.
August 22, 2026: Claude Opus 5: Anthropic's safest model at half the price
Anthropic released Claude Opus 5, claiming it is its safest model ever while costing half the price of Opus 4. This is a significant improvement in both safety and economics. Read the full article.
August 22, 2026: DeepSeek V4 Flash Vision: multimodal AI takes on Claude
DeepSeek released V4 Flash Vision, a multimodal model that directly challenges Claude's vision and reasoning capabilities. It supports both text and image input, expanding beyond DeepSeek's previous text-only offerings. Read the full article.
What to Watch
- Open-weight trend: DeepSeek, Tencent, and Z.AI are all competing through open-weights. This is reshaping the AI landscape because open models are catching up to proprietary ones in capability. See the latest benchmarks on BenchLM.
- Anthropic's position: Despite losing spending ground to OpenAI, Anthropic is still releasing top-tier models like Claude Opus 5 and planning a major IPO in October.
- Chinese AI companies: Tencent Hy4 and Z.AI Ox Alpha signal that Chinese companies are no longer playing catch-up. They are leading in open-weight releases.
Last updated: August 31, 2026. This tracker is updated in place as new developments break. Follow our AI coverage for full articles. For the latest AI model benchmarks, visit BenchLM's leaderboard.