Claude Opus 5.5 — released 22 September 2026, the first model in Anthropic's Claude 5.5 family and the new default Opus. Anthropic reports Claude Fable 5.1-level performance at roughly 40% lower running cost than Opus 5, a 1M-token context window, and strengthened cybersecurity safeguards. API id: claude-opus-5-5.

GPT-6 Sol is a lower-cost model in OpenAI's GPT-6 family, released on 22 September 2026 alongside GPT-6 Luna and trained with methods similar to the flagship GPT-6 Astra. It is priced at $2 per million input tokens and $10 per million output tokens, and OpenAI claims $0.27 per task on AutomationBench. Available in the API at launch; ChatGPT Plus and Pro users get it in Work and Codex rather than regular chat.

GPT-6 Luna is the cheapest model in OpenAI's GPT-6 family, released on 22 September 2026 alongside GPT-6 Sol and trained with methods similar to the flagship GPT-6 Astra. It is priced at $0.10 per million input tokens and $0.50 per million output tokens — roughly a twentieth of Sol — and was available in the API at launch.

Grok 4.7 is the large language model xAI — now operating as SpaceXAI inside SpaceX — released on September 21, 2026 as the successor to Grok 4.6. It is a larger base model aimed at coding and knowledge work, priced at the same $2 per million input and $6 per million output tokens as Grok 4.6, and it was available in GitHub Copilot at launch; early independent benchmarks placed it well behind Claude and GPT-6.

Qwen-Image 2.1 is an open-weight image model from Alibaba's Qwen team, released in September 2026, that handles both image generation and image editing in a single 7-billion-parameter model. It can output images with native transparency (RGBA), and Alibaba claims it beats closed models on image generation despite its size; the weights are released under a non-commercial, research-only license.

Koa is Salesforce's first CRM reasoning model, unveiled at Dreamforce on September 15, 2026. Built with NVIDIA by post-training NVIDIA's open-weight Nemotron 3 Super on synthetic scenarios drawn from nearly three decades of Salesforce CRM knowledge, it is designed to handle complex, multi-step sales, marketing and customer-support workflows and to take actions across CRM applications in Agentforce. It arrived three weeks after Salesforce made Anthropic's Claude a default model, a move analysts tied to inference costs.

Jev is an AI model from TypeSafe AI, a startup whose founders include researchers who worked on ChatGPT, launched on September 15, 2026 as the company emerged from stealth. Rather than generating text, Jev is a "System One" model that returns typed, calibrated decisions that software can act on directly, and it is pitched as a fast, low-cost alternative to calling a large language model for routine decisions inside applications.

Gemini 3.8 Live is Google's real-time voice model, released on September 15, 2026 together with a Gemini 3.8 Live Extended Thinking variant. Google calls them its most advanced live dialogue models: they run tool and API calls in the background while the conversation continues, understand what the camera sees and support 97 languages. Both are available in the Gemini API, Google AI Studio and Gemini Enterprise, and Gemini 3.8 Live now powers Search Live in the Google app.

Suno v6 is the AI music generation model Suno released on September 9, 2026 — its first model built with support from the record industry, following partnerships with Warner Music and BMG. Suno says v6 was trained from the ground up on a new, licensed dataset rather than the music used for earlier versions, and it replaces the retiring v4.5 and v5 models as the company faces a wave of copyright lawsuits.

ChatGPT Images 2.5 is OpenAI's image generation model released on September 8, 2026, the successor to GPT Image 2 and also referred to as GPT Image 2.5. It brings faster generation, sharper detail and more precise editing that changes only the requested part of an image, and launched alongside ChatGPT Sketch, which turns rough drawings into detailed images.

Gemini 3.8 Flash is Google's fast, low-cost Gemini model released on September 2, 2026 — its third Flash model in six weeks, arriving shortly after Gemini 3.7 Flash. Google says it "works harder" by performing more reasoning steps on complex tasks, which can raise costs. It shipped alongside Gemini 3.8 Flash Cyber, a cybersecurity variant of the same underlying model that differs only in its safety mitigations.

DeepSeek V4.1 Flash is DeepSeek's September 2026 update to its fast, low-cost V4 Flash model. It supports a 1-million-token context window and adds an FP4 KV cache and cross-layer attention reuse to cut the memory cost of long-context, agent-style workloads; early testers reported it reaching about 98% of GPT-6 Astra's score on the OpenDesign Arena at roughly 1.4% of the cost.

AlphaGenome is Google DeepMind's deep-learning model for predicting how DNA sequences regulate gene expression and how single-letter variants can disrupt that regulation, first released in June 2025. In September 2026 DeepMind published AlphaGenome Atlas, a roughly 1-petabyte precomputed map of the predicted molecular effects of all 9 billion possible single-letter changes in the human genome, with an impact score for every variant.

GPT-6 Astra is OpenAI's flagship model, released on September 3, 2026 and positioned as a computer-use and coding model rather than a chat model. It ships with a ~1.05M-token context window, reports 72.6% on OSWorld V2-Offline, and is the first OpenAI model to reach the Critical cybersecurity capability level under the company's Preparedness Framework — a milestone OpenAI leadership has described as the start of an AGI era.

Claude Fable 5.1 is Anthropic's September 1, 2026 upgrade to Fable 5, state of the art on coding, knowledge work and long-running agentic problem-solving — 52.6% on Terminal-Bench-Science 0.1 (vs 24.7% for Fable 5), 73.4% on CursorBench 3.2.0 at max effort and 65.0% on Humanity's Last Exam with tools. Generally available on Claude apps, the API, AWS, Google Cloud and Microsoft Azure; a 75% cut in cache-read pricing (to $0.25 per million tokens) makes typical workloads about 25% cheaper than Fable 5.

Claude Mythos 5.1 is the same underlying model as Claude Fable 5.1, released September 1, 2026, but with safeguards lifted for verified work in cybersecurity and the life sciences — including a 60% reduction in false-positive refusals in the cyber domain and advanced protein-design capability (high-affinity binder hit rates near 50%). It is not generally available: access runs only through Anthropic's Cyber Verification Program for defensive security professionals and the Life Sciences Verification Program for US-based researchers.

Wan 3.0 is the third generation of Alibaba's Wan video-generation model family, opened as an invite-only beta on QwenCloud on August 6, 2026. It targets native 4K output, single-shot clips up to 30 seconds with generated audio, and character consistency across multi-shot scenes. Unlike Wan 2.1/2.2, which shipped Apache-2.0 open weights, Wan 3.0 is so far a closed beta and API — no public weights have been released.

Ox Alpha is an anonymous “stealth” AI model that appeared on OpenRouter on August 20, 2026 — a free-preview reasoning model aimed at coding and agentic work, with a 1,048,576-token context window, a 131,072-token output cap and text, image and video input. No lab has claimed it; community fingerprinting (tokenizer probes, API error strings, video-token budgets) points at Z.ai’s GLM family.

Gemini 3.7 Flash is Google's fast, low-cost Gemini tier released in August 2026, aimed at coding and agentic workloads at $0.75 per million input tokens — arriving three weeks after the previous Flash release.

GLM-5.3 is the open-weight large language model Z.ai (Zhipu AI) launched on August 13, 2026 as the successor to GLM-5.2, positioned as a challenger to Anthropic and OpenAI in coding. Z.ai said it would publish the weights within two weeks of launch, and a smaller GLM-5.3-Flash variant followed. Its bug-finding ability drew attention from security researchers, and OpenAI's Greg Brockman warned it could accelerate the cyber threat landscape.

DeepSeek-V4-Pro — the flagship tier of DeepSeek's trillion-scale V4 family, a MoE model aimed at frontier reasoning, coding and agentic workloads above the cheaper V4 Flash tier.

Grok 4.6 — xAI's frontier large language model and successor to Grok 4.5, a 500K-context model tuned for long-running agents, coding and real-time knowledge via X integration.

Meta's Muse Glimmer — a 30B open-weights, multimodal agentic model built for always-on local agent workflows, small enough to run on a single consumer GPU.

Meta's Muse Spark 1.2 — the August 2026 update to Meta's paid Muse Spark foundation model, powering the Muse Code terminal agent with higher coding and Terminal-Bench scores at a low per-task price.

NVIDIA's Nemotron 4 — the next major generation of the Nemotron open model family, following the Nemotron 3 and Nemotron 3.5 releases, aimed at reasoning-heavy and agentic enterprise AI workloads.

NVIDIA's Nemotron 3.5 Lightning — the speed-optimized tier of the Nemotron 3.5 open model family, built for fast, accurate specialized task execution in long-running agentic workflows.

WeatherNext is Google DeepMind's family of machine-learning weather forecasting models, which predict global atmospheric conditions and cyclone tracks up to 15 days ahead. WeatherNext 2 generates each forecast scenario in under a minute on a single TPU — far faster than traditional physics-based numerical weather prediction — and powers weather features across Google Search, Gemini and Pixel.

Alibaba Cloud's latest Qwen 3-series flagship LLM, the successor to Qwen 3.7 with improved reasoning, coding and agentic capabilities.

NVIDIA's Nemotron 3 Nano — the compact, efficiency-focused tier of the Nemotron 3 open model family, including the Nano Omni multimodal variant for long-context document, audio and video agents.

GPT-5.6 Sol — the flagship tier of OpenAI's GPT-5.6 family, positioned as its highest-capability reasoning model and benchmarked against Claude Opus 5 and Fable 5.

GPT-5.6 Luna — the low-cost, high-throughput tier of OpenAI's GPT-5.6 family, whose price cuts drove the 2026 inference price war.

GPT-5.6 Terra — the mid tier of OpenAI's GPT-5.6 family, sitting between Luna and Sol on price and capability, generally available via the OpenAI API and Amazon Bedrock.

DeepSeek-V4-Flash — the fast, low-cost tier of DeepSeek's V4 family, released at $0.28 per million tokens and upgraded (0731) with stronger agentic and coding performance.

Udio's text-to-music model generates complete songs — vocals, lyrics and backing instrumentation — from a natural-language prompt.

Granite is IBM's family of open, Apache 2.0 licensed language, vision and embedding models aimed at enterprise workloads.

Claude Opus 5 — Anthropic's most capable model in the Claude 5 family, the frontier successor to Opus 4.8 for the hardest reasoning, coding and agentic work.

Mureka V9.5 is the AI music generation model from Mureka, the music platform of China's Kunlun Tech, introduced in July 2026. Kunlun pitches it as making songs that sound less machine-made by using "reflective reasoning" and agentic song creation, and in August 2026 Mureka launched it internationally alongside a second music model called O3.

Seedance 2.5 is ByteDance's AI video generation model, released in July 2026 as the successor to Seedance 2.0. It generates single-take clips of up to 30 seconds, accepts up to 50 reference inputs and adds 3D camera blocking for directing shots; its API went live in July 2026, and it is also offered through third-party creative platforms such as Krea.

DeepSeek V4 — DeepSeek's frontier model with a million-token context window built for long-running agentic workloads.

GLM-5.2 — Zhipu AI's flagship GLM model tuned for long-horizon, multi-step agentic tasks.

Step 3.7 — StepFun's enterprise-ready multimodal model (including the Step 3.7 Flash tier), optimised for NVIDIA GPU inference.

Inkling — Thinking Machines Lab's first model trained from scratch: a 975B-parameter open-weights multimodal Mixture-of-Experts with 41B active parameters and controllable thinking effort, released under Apache 2.0 in July 2026.

Kimi K3 — Moonshot AI's next-generation Kimi model, reported to close the gap with Anthropic's Opus 4.8 and shipped as one of the largest open-source frontier models.

Amazon Nova — Amazon's family of foundation models on AWS Bedrock, spanning text, vision and the Nova Act agentic browser model.

MiniMax M3 — MiniMax's long-context reasoning and agentic model, positioned for open-weight deployment on accelerated infrastructure.

Gemini 3.1 Pro — Google's frontier Gemini model for complex reasoning tasks, benchmarked against GPT-5.4, Claude Opus 4.6 and Grok.

Gemini Omni Flash — the fast, low-cost tier of Google's Gemini Omni video model, aimed at conversational enterprise video generation via the API.

Nano Banana 2 Pro — the flagship tier of Google's Nano Banana 2 image family, powered by Gemini 3.1 Flash Image.

Grok 4.5 — xAI's frontier large language model, the successor to Grok 4, with stronger reasoning, coding and real-time knowledge via X integration.

GPT-Live-1 is OpenAI's full-duplex voice model, which can listen and speak at the same time. OpenAI introduced it as GPT-Live on July 8, 2026, powering ChatGPT's upgraded voice mode, and opened it to developers in the API on September 10, 2026 at $0.05 per minute for the voice layer. The model handles the conversation itself — turn-taking and back-channel cues — and hands the reasoning to a separate model the developer chooses.

Watermelon — the codename for Meta's upcoming frontier AI model, reported to match OpenAI's GPT-5.5 on key benchmarks. Meta's bid to reclaim ground in the frontier LLM race under Alexandr Wang's superintelligence group.

Nano Banana 2 Lite — Google's fastest, cheapest tier of its Nano Banana 2 (Gemini image) family, built for high-volume, low-latency image generation.

Claude Sonnet 5 — Anthropic's balanced mid-tier model in the Claude 5 family, pairing strong reasoning and coding with fast, cost-efficient responses.

PixVerse V6 — PixVerse's 2026 flagship video generation model, advancing toward real-time, cinematic control across creative and agentic workflows.

PixVerse R1 — PixVerse's real-time video 'world model' with live input, subject priority, shared worlds and personalized avatars.

Meituan's 1.6-trillion-parameter open-source agentic coding model with a 1M-token context window (June 2026). The first trillion-parameter model claimed to be fully pre-trained AND served on domestic Chinese AI chips.

Mistral AI's document-intelligence (OCR) model (June 2026). Structure-aware extraction with bounding boxes, block classification and confidence scores across 170 languages; self-hostable in a single container. Tops OlmOCRBench; feeds RAG, agentic and enterprise-search pipelines.

OpenAI's GPT-5.6 family (previewed 2026) — tiered Sol, Terra and Luna models with new reasoning modes; reportedly beats Anthropic's Mythos 5. Initial rollout was limited at US government request.

Recraft's most advanced text-to-image model (V4.1) — design-grade image generation with strong visual taste, brand styles, and precise control.

Sakana AI model that coordinates and orchestrates multiple models, matching frontier models on some benchmarks.

The GLM (General Language Model) family from Zhipu AI / Z.ai — open-weight Chinese LLMs (GLM-4.5/4.6/5/5.1/5.2 plus GLM-OCR, Air, Flash and Turbo variants) known for strong coding and long-context performance under permissive (MIT) licenses.

NVIDIA's Nemotron 3 Ultra — the largest, highest-capability model in the Nemotron 3 open model family, tuned for advanced reasoning, agentic workflows and enterprise AI.

Anthropic's Mythos-class multimodal Claude model (text, vision, code) made safe for general use — strong at software engineering, knowledge work, long-context reasoning, scientific research and protein design, with safeguards that fall back to Claude Opus 4.8 for sensitive domains. Available via the Claude API and claude.ai. Priced at $10 / $50 per million input / output tokens.

The same underlying multimodal Claude model as Fable 5, but with safeguards lifted in some domains (e.g. cybersecurity, biology), offered to vetted users through Anthropic's Project Glasswing trusted-access program. Text, vision and code. Priced at $10 / $50 per million input / output tokens.

Google's 12-billion-parameter open model in the Gemma 4 family — a compact, efficient multimodal LLM designed to run on a single GPU.

Anthropic's Claude Sonnet 4.8, launched alongside Opus 4.8 — brings Opus-class quality into the mid tier with the new Dynamic Workflows tool and improved vision workflows.

Anthropic's flagship Claude Opus 4.8 — the successor to Opus 4.7 with further improvements in advanced reasoning, coding and agentic capabilities.

Alibaba Cloud's latest Qwen 3-series flagship LLM, the successor to Qwen 3.6 with improved reasoning, coding and agentic capabilities.

NVIDIA Cosmos — world foundation models that generate physics-aware synthetic data and reasoning for physical AI and robotics.

Google DeepMind's mathematical-reasoning model that formally proves theorems; the AlphaProof Nexus version tackles Erdős problems.

Google DeepMind's embodied-reasoning Gemini model for real-world robotics tasks.

NVIDIA's foundation model for humanoid robots (Isaac GR00T), enabling generalist embodied skills.

Physical Intelligence's Vision-Language-Action (VLA) models for general robot control (π0, π0-FAST, π0.6).

Google's fast, cost-efficient Gemini model tier, announced at Google I/O 2026.

OpenAI's image generation model, successor to DALL·E, integrated into ChatGPT and the API.

ByteDance's unified model for image and video understanding, generation and editing.

OpenAI's GPT-5.5 — the default ChatGPT model (Instant), with Pro and Thinking variants for paid plans. Lower hallucination in law, medicine and finance, and stronger STEM reasoning. A GPT-5.5-Cyber variant targets cybersecurity.

Google's multimodal model family for video generation and editing, announced at Google I/O 2026. Gemini Omni Flash creates and edits high-quality video from text, image, audio and video inputs with physics-aware generation, conversational editing, digital avatars and SynthID watermarking.

OpenAI's speech-to-text AI model

Moonshot AI's flagship 1T-parameter open-weight LLM featuring 262K context window, long-horizon coding with up to 300 sub-agent swarms and 4,000 coordinated steps. Outperforms GPT-5.4 and Claude Opus 4.6 on SWE-Bench Pro (58.6). Supports multimodal input including vision.

Opus 4.7 is a notable improvement on Opus 4.6 in advanced software engineering, with particular gains on the most difficult tasks.

Meta open-source AI video generation model for creating short video clips from text and image prompts.

Google DeepMind's most capable open model family. Available in 4 sizes (E2B, E4B, 26B MoE, 31B Dense) with advanced reasoning, agentic workflows, vision, audio, 256K context, 140+ languages. Apache 2.0 license. Runs on devices from phones to H100 GPUs.

Most cost-effective video generation model is now available to developers in the Gemini API.

By far the most powerful AI model Anthropic ever developed.

Lyria 3 helps you express, explore, and experiment with high-fidelity music, using prompts to create tracks with natural flow from note to note.

Most capable and efficient frontier model for professional work.

Hybrid reasoning model with superior intelligence for agents, featuring a 1M context window

Claude Opus 4.6 is state-of-the-art across a wide range of coding and agentic capabilities.

Grok is an AI assistant built by xAI. Chat, create images, write code, and get real-time answers from the web and X

Claude Haiku 4.5 is our fastest, most cost-efficient model, matching Sonnet 4’s performance on coding, computer use, and agent tasks.

Seedance 2.0 adopts a unified multimodal audio-video joint generation architecture that supports text, image, audio, and video inputs, leading to the most comprehensive multimodal content reference and editing capabilities in the industry.