Last Updated: November 18, 2026
Summary: The major LLMs in late 2026 are OpenAI's GPT-6 Astra, Anthropic's Claude Fable 5.2 and Opus 5.5, Google's Gemini 4 Argon and Carbon, xAI's Grok 4.7 and Meta's Muse Wonder on the closed side, plus open-weight models such as DeepSeek-V4.1-Flash, Qwen3.8, Kimi K3.1 Pro and Mistral Large 4. API prices for flagship models run from about $0.30 to $10 per million input tokens, and most flagships now accept around 1 million tokens of context.
New large language models ship almost every week, so older "top 10" lists go stale fast. This list of LLMs groups every major model by company, with release dates, weights, context windows and official API prices, checked against vendor documentation in November 2026.
New to the topic? Our guide to what an LLM is covers the basics.
Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.
In this guide
List of LLMs in 2026: Major Models at a Glance
The table lists the current flagship LLM from each major developer. "Open" means the weights are downloadable; "closed" means access only through the company's apps or API. Prices are standard API rates per 1 million tokens (input / output) from each company's pricing page or launch post.

The major LLMs at a glance, from closed APIs to open weights. Photo: Planet Volumes / Unsplash+
Company | Flagship LLM | Released | Weights | Context window | API price per 1M tokens (input / output) |
|---|---|---|---|---|---|
OpenAI | GPT-6 Astra | Sep 3, 2026 | Closed | 1,050,000 | $10 / $50 |
Anthropic | Claude Fable 5.2 | Nov 9, 2026 | Closed | 1M | $8 / $40 |
Anthropic | Claude Opus 5.5 | Sep 22, 2026 | Closed | 1M | $4 / $20 |
Gemini 4 Argon | Sep 30, 2026 | Closed | 1,048,576 | $2 / $10 until Nov 30, 2026, then $4 / $20 | |
Gemini 4 Carbon | Nov 10, 2026 | Closed | Not stated | $2 / $10 until Nov 30, 2026, then $4 / $20 | |
xAI | Grok 4.7 | Sep 21, 2026 | Closed | 500K | $2 / $6 (prompts under 200K) |
Meta | Muse Wonder | Oct 12, 2026 | Closed (API only) | Not stated | $4 / $20 (reported by TechCrunch) |
DeepSeek | DeepSeek-V4.1-Flash | Sep 10, 2026 | Open | 1M | $0.30 / $1.20 peak; half off-peak |
Alibaba | Qwen3.8 (2.4T-A95B) | Aug 12, 2026 | Open | Not stated | Not published |
Moonshot AI | Kimi K3.1 Pro | Oct 23, 2026 (API) | Open | 1,048,576 | ¥13 / ¥68 (yuan, Moonshot's posted rate) |
Mistral AI | Mistral Large 4 | Oct 6, 2026 (preview) | Open | Not stated | $0.70 / $2.10 |
Microsoft | MAI-Thinking-1 | Aug 12, 2026 (public preview) | Not disclosed | 256K | Not published |
Amazon | Nova 2 Lite | Dec 2, 2025 | Closed (Bedrock only) | 1M | See Bedrock pricing |
Cohere | Command A+ | May 20, 2026 | Open (Apache 2.0) | 128K | Not published in docs |
A 1 million token context window is now standard for flagships from OpenAI, Anthropic, Google, DeepSeek and Moonshot. The price gap is enormous: GPT-6 Astra input costs about 33 times more than DeepSeek-V4.1-Flash at peak rates ($10 versus $0.30).
Which Are the Major LLMs From Closed-Model Companies?
The major closed LLMs come from OpenAI, Anthropic, Google, xAI and Meta, plus Microsoft and Amazon for their cloud customers.
OpenAI: GPT-6 family
OpenAI's model catalog lists four flagship models. OpenAI's developer docs describe GPT-6 Astra as its most capable model, GPT-6.1 Sol as close to Astra at lower cost, GPT-6.1 Terra as a fast model for visual reasoning and writing, and GPT-6 Luna as the most efficient option for high-volume work. According to the OpenAI changelog, Astra launched September 3, 2026, Sol and Luna followed on September 22, GPT-6.1 Sol on September 29 and GPT-6.1 Terra on October 26.
OpenAI's model pages list a 1,050,000 token context window and 128,000 max output tokens for Astra, 6.1 Sol and Luna. API rates are $10 / $50 for Astra, $2 / $10 for GPT-6.1 Sol and $0.10 / $0.50 for Luna. OpenAI's open-weight models are gpt-oss-120b and gpt-oss-20b, released in August 2025 under Apache 2.0. Full rates and discounts are in our OpenAI API pricing guide.
Anthropic: Claude Fable, Opus, Sonnet and Haiku
Anthropic's current Claude lineup has four models, all with a 1M token context window and 128K max output, per Anthropic's models overview:
Claude Fable 5.2 ($8 / $40): for demanding reasoning and long-horizon agent work, released November 9, 2026
Claude Opus 5.5 ($4 / $20): Anthropic's recommended default for most workloads, released September 22, 2026
Claude Sonnet 5.5 ($2 / $10): released September 28, 2026
Claude Haiku 5.5 (from $0.10 / $0.50): released October 7, 2026
Anthropic also runs Claude Mythos 5.2, which its launch post says is the same underlying model as Fable 5.2 with fewer cybersecurity and biology restrictions, limited to vetted organizations. See our Claude API pricing breakdown for caching and batch rates.
Google: Gemini 4 Argon and Carbon
Google introduced Gemini 4 Argon on September 30, 2026, with output of up to 1 million tokens, up from 64K. The Gemini API model page lists Argon as stable with a 1,048,576 token context window.
Gemini 4 Carbon arrived on November 10, 2026. Google calls Carbon its strongest model, with frontier cybersecurity performance, aimed at the hardest multi-step coding and research tasks, while saying Argon "remains the ideal choice for the majority of software engineering and knowledge work." Both carry an introductory API price of $2 / $10 through November 30, 2026, rising to $4 / $20 from December 1, per Google's Carbon announcement and Vertex AI pricing page. Our Gemini API pricing guide covers the rest of the lineup.
xAI: Grok 4.7
xAI's model documentation describes Grok 4.7 as its most capable model, with a 500K context window and pricing of $2 input and $6 output per million tokens for prompts under 200K tokens. xAI announced Grok 4.7 on September 21, 2026, with access through the Grok API, Grok Build and Cursor.
Meta: Muse Wonder and Muse Spark
Meta's newer Muse models come from Meta Superintelligence Labs. DeepLearning.AI's The Batch described April 2026's Muse Spark as a proprietary model and Meta's pivot away from its open-weights Llama strategy. Muse Spark 1.1 followed on July 9 with a 1 million token context window, per Meta's announcement. Meta's most capable model, Muse Wonder, launched October 12, 2026 through the Meta Model API; TechCrunch reported its price at $4 / $20.
Microsoft and Amazon
Microsoft's in-house reasoning model, MAI-Thinking-1, is in public preview in Microsoft Foundry with a 256K token window, according to Microsoft AI. Amazon's Nova 2 Lite, its multimodal model on Bedrock, launched December 2, 2025, with a 1M token context window per its Bedrock model card.
What Are the Best Open-Weight LLMs Right Now?
The leading open-weight LLMs in late 2026 are DeepSeek-V4.1-Flash, Alibaba's Qwen3.8, Moonshot's Kimi K3.1 Pro and Mistral Large 4. Google's Gemma 4 and OpenAI's gpt-oss cover the smaller end.
DeepSeek-V4.1-Flash
DeepSeek released DeepSeek-V4.1-Flash on September 10, 2026, describing it as a 552B-parameter mixture-of-experts model with weights on Hugging Face. DeepSeek's pricing page lists a 1M context window, 384K max output and peak rates of $0.30 input (cache miss) and $1.20 output, with off-peak rates at half price. DeepSeek is phasing out V4-Pro: since September 14, its API routes V4-Pro requests to V4.1-Flash until V4.1-Pro launches. Our DeepSeek statistics page tracks the company's growth.
Alibaba Qwen3.8
Alibaba's Qwen3.8 repository shows Qwen3.8-2.4T-A95B reaching Hugging Face on August 12, 2026, followed by Qwen3.8-27B on August 14. Alibaba's most recent proprietary model is Qwen3.7-Max, announced May 21, 2026.
Moonshot AI Kimi K3.1 Pro
Moonshot's Kimi K3 launched in July 2026 as a 2.8 trillion parameter model with 104B active parameters and a 1,048,576 token context, per its Hugging Face model card. Kimi K3.1 Pro, which Moonshot's announcement calls a post-trained version of K3, reached the Kimi API on October 23, 2026 at ¥13 input and ¥68 output per million tokens, and TechNode reported on November 5 that Moonshot had released its weights on Hugging Face.
Mistral Large 4
Mistral launched Mistral Large 4 in public preview on October 6, 2026 and later released the weights of this 1 trillion parameter model under its own open-weight license, according to Mistral's announcement. API pricing is $0.70 input and $2.10 output per million tokens. Read more in what is Mistral AI.
Gemma 4, gpt-oss and Command A+
Google released Gemma 4 on April 2, 2026 in four sizes (E2B, E4B, 26B MoE and 31B dense) under Apache 2.0, with up to 256K context on the larger models; gpt-oss supports 128K. Cohere's Command A+, released May 20, 2026, is a 218B-parameter MoE model with 128K input, listed under Apache 2.0.
What Are the Newest LLM Releases?
The newest major LLMs are Google's Gemini 4 Carbon (November 10, 2026) and Anthropic's Claude Fable 5.2 (November 9, 2026). Here are the major LLM releases from the last 90 days, dated from official announcements:
Date (2026) | Model | Company | Weights |
|---|---|---|---|
Nov 10 | Gemini 4 Carbon | Closed | |
Nov 9 | Claude Fable 5.2 and Mythos 5.2 | Anthropic | Closed |
Oct 26 | GPT-6.1 Terra | OpenAI | Closed |
Oct 23 | Kimi K3.1 Pro (API) | Moonshot AI | Open |
Oct 12 | Muse Wonder | Meta | Closed |
Oct 7 | Claude Haiku 5.5 | Anthropic | Closed |
Oct 6 | Mistral Large 4 (preview) | Mistral AI | Open |
Sep 30 | Gemini 4 Argon | Closed | |
Sep 29 | GPT-6.1 Sol | OpenAI | Closed |
Sep 28 | Claude Sonnet 5.5 | Anthropic | Closed |
Sep 22 | Claude Opus 5.5; GPT-6 Sol and GPT-6 Luna | Anthropic; OpenAI | Closed |
Sep 21 | Grok 4.7 | xAI | Closed |
Sep 10 | DeepSeek-V4.1-Flash | DeepSeek | Open |
Sep 3 | GPT-6 Astra | OpenAI | Closed |
That is 14 release dates in about ten weeks, and Anthropic alone shipped four new models in seven weeks.
Which Companies Make LLMs?
The leading LLM companies are OpenAI, Anthropic, Google, Meta, xAI, DeepSeek, Alibaba, Moonshot AI, Mistral AI and Cohere, with Microsoft and Amazon building models for their own clouds.

Most major LLMs come from a small group of AI labs. Photo: Getty Images / Unsplash+
LLM company | Model family | Approach | Where developers access it |
|---|---|---|---|
OpenAI | GPT-6, gpt-oss | Mostly closed | OpenAI API, Microsoft Azure, AWS Bedrock |
Anthropic | Claude | Closed | Claude Platform, AWS, Google Cloud, Microsoft Azure |
Gemini, Gemma | Closed and open | Gemini API in Google AI Studio | |
xAI | Grok | Closed | Grok API, Cursor |
Meta | Muse (formerly Llama) | Proprietary (Muse) | Meta Model API |
DeepSeek | DeepSeek V4 | Open | DeepSeek API, Hugging Face |
Alibaba | Qwen | Open and closed | Hugging Face, Alibaba Cloud Model Studio |
Moonshot AI | Kimi | Open | Kimi API, Hugging Face |
Mistral AI | Mistral, Ministral | Mostly open | Mistral Studio, Hugging Face |
Microsoft | MAI | Not disclosed | Microsoft Foundry |
Amazon | Nova | Closed | Amazon Bedrock |
Cohere | Command | Open (Apache 2.0) | Cohere API |
These are also called foundation models. Our AI market share report shows which companies lead in usage.
Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.
How Much Do the Top LLMs Cost to Run?
The same workload can cost $200 on one LLM and under $6 on another. We priced a monthly job of 10 million input and 2 million output tokens at each company's standard rate, before caching or batch discounts:
Model | Input cost | Output cost | Total |
|---|---|---|---|
GPT-6 Astra | $100 | $100 | $200 |
Claude Fable 5.2 | $80 | $80 | $160 |
Claude Opus 5.5 | $40 | $40 | $80 |
Gemini 4 Argon (after intro price) | $40 | $40 | $80 |
GPT-6.1 Sol | $20 | $20 | $40 |
Claude Sonnet 5.5 | $20 | $20 | $40 |
Grok 4.7 (prompts under 200K) | $20 | $12 | $32 |
Mistral Large 4 | $7 | $4.20 | $11.20 |
DeepSeek-V4.1-Flash (peak) | $3 | $2.40 | $5.40 |
Claude Opus 5.5 and Gemini 4 Argon land at the same $80 once Argon's introductory pricing ends on November 30 (until then, Argon costs $40). GPT-6.1 Sol and Claude Sonnet 5.5 cost $40, a fifth of GPT-6 Astra. DeepSeek-V4.1-Flash at off-peak rates would run this job for $2.70, about 1.4% of the Astra bill, though companies with data residency rules should check where a model is hosted first.
Open vs Closed LLMs: Which Should You Pick?
Pick a closed LLM for the strongest results with no infrastructure to manage, and an open-weight LLM to control where the model runs or cut costs at high volume.
For the hardest reasoning and agent work: GPT-6 Astra, Claude Fable 5.2 or Gemini 4 Carbon. Expect to pay $8 to $10 per million input tokens for Astra and Fable.
For everyday business work: Claude Opus 5.5 or Sonnet 5.5, GPT-6.1 Sol or Gemini 4 Argon.
For high-volume, cost-sensitive tasks: GPT-6 Luna at $0.10 / $0.50, Claude Haiku 5.5 or DeepSeek-V4.1-Flash.
For self-hosting and data control: Qwen3.8, Kimi K3.1 Pro, Mistral Large 4 or DeepSeek's open weights; smaller teams can try Gemma 4.
Sameer's take: I pay for Claude and ChatGPT, and Claude is my favourite of all the AI tools.
How Many LLM Models Are There?
No official count exists, because thousands of open-weight variants are published on Hugging Face. BenchLM's directory lists 1,031 models. Even a single vendor's catalog runs long: Anthropic's pricing page lists 21 Claude models, of which Anthropic recommends only four as current.

Older model versions keep running long after new ones ship. Photo: Getty Images / Unsplash+
Frequently Asked Questions
What are examples of LLMs?
Examples of LLMs include OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5, Google's Gemini 4 Argon, xAI's Grok 4.7 and Meta's Muse Wonder. Open-weight examples include DeepSeek-V4.1-Flash, Alibaba's Qwen3.8, Moonshot's Kimi K3.1 Pro and Mistral Large 4.
What are the top 5 LLM models?
As of November 2026, the top five LLMs by vendor positioning are GPT-6 Astra, Claude Fable 5.2, Gemini 4 Carbon, Gemini 4 Argon and Claude Opus 5.5. Each company describes these as its most capable or recommended models, and all support context windows of around 1 million tokens where stated.
What is the newest LLM?
The newest major LLM is Google's Gemini 4 Carbon, announced November 10, 2026, one day after Anthropic released Claude Fable 5.2 on November 9, 2026. TechNode reported on November 5 that Moonshot had released the open weights of Kimi K3.1 Pro.
Are all LLMs GPTs?
No. GPT is OpenAI's brand name for its models. Claude, Gemini, Grok, Muse, DeepSeek, Qwen, Kimi and Mistral are all LLMs from other companies, each with its own model family and pricing.
Which LLM is the best?
No single LLM is best for every task. GPT-6 Astra and Claude Fable 5.2 target the hardest work, while Claude Sonnet 5.5 and GPT-6.1 Sol offer strong results for about a fifth of the cost, and DeepSeek-V4.1-Flash is the cheapest major option.
The Bottom Line
The list of LLMs worth knowing in 2026 is shorter than it looks: GPT-6, Claude, Gemini 4, Grok and Muse on the closed side, and DeepSeek, Qwen, Kimi and Mistral leading open weights. Almost every major lab shipped a new flagship between September and November 2026, so treat any model list as a snapshot.
For most teams, test a mid-tier model such as Claude Sonnet 5.5 or GPT-6.1 Sol first, then move up to a flagship only where it earns the extra cost.
Related Articles
What Is an LLM?: how large language models work.
What Is a Foundation Model?: the models businesses build on.
AI Market Share 2026: which AI companies lead in usage.
OpenAI API Pricing: token rates for every GPT model.
Claude API Pricing: token rates for every Claude model.
Gemini API Pricing: Google's model rates explained.
What Is Mistral AI?: Europe's leading AI lab.
DeepSeek Statistics 2026: users, growth and costs.
⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.
Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.
