Large Language Models

Free for non-commercial and <$1M revenue. API: $0.065 per image (Large). Enterprise license for $1M+ revenue companies.

GPT-5.2 operates under a strict proprietary license, with access provided exclusively through OpenAI's API and its ChatGPT consumer products. The API pricing for the standard GPT-5.2 model is set at $1.75 per million input tokens and $14.00 per million output tokens. OpenAI offers significant cost-saving mechanisms, including a 90% discount for cached inputs ($0.175 per million tokens) and a 50% discount for asynchronous tasks via the Batch API. The flagship GPT-5.2 Pro model is priced at a sig

Gemini 3 Pro is a proprietary model available through Google's developer APIs. During its preview phase, Google has implemented a tiered pricing structure based on the length of the context used. For context lengths up to 200,000 tokens, the API pricing is $2.00 per million input tokens and $12.00 per million output tokens. For contexts exceeding 200,000 tokens, the price increases to $4.00 per million input tokens and $18.00 per million output tokens. Google has a history of reducing prices on

The Claude 4.5 models are governed by a proprietary license and are accessible via Anthropic's API and through major cloud platforms. The pricing is structured per million tokens (MTok). For Claude Opus 4.5, the rate is $5.00 per MTok for input and $25.00 per MTok for output. For Claude Sonnet 4.5, the pricing for its standard 200K context window is $3.00 per MTok for input and $15.00 per MTok for output. When using the extended 1M context window, the rates for Sonnet 4.5 increase to $6.00 per

Starter: Free (150 AI credits/day). Professional: $16/month (3,000 AI credits). Organization: $55/month (3,500 credits). Enterprise: $90/month (4,250 credits). Figma AI module: $16.

Free: 1,000 credits/month. Core: $9/month (10,000 credits). Pro: $16/month (10,000 credits). Teams: $29/month (10,000 credits). Enterprise: Custom. Credit-based system.

Free: 150 daily tokens. Apprentice: $12/month (8,500 tokens). Artisan: $30/month (25,000 tokens + unlimited relaxed). Maestro: $60/month (60,000 tokens + unlimited). API: $9-$299/month.

Free: 50 credits/day (~10 songs, non-commercial). Pro: $8-$10/month (~500 songs). Premier: $24-$30/month (~2,000 songs). Commercial licensing with paid plans.

Free: Limited AI trial (20 responses). Plus: $12/month ($10 annual). Business: $24/month ($20 annual, includes full AI). Enterprise: Custom (includes full AI).

Free: 10 credits/day + 100 monthly. Standard: $10/month ($96 annual, 1,200 credits). Pro: $30/month ($288 annual, 4,800 credits). Commercial use rights included.

Hobby: Free (limited). Pro: $20/month (500 premium requests + $20 API credit). Ultra: $200/month. Teams: $40/user/month. Enterprise: Custom.

Free: 2,000 completions + 50 chats/month. Pro: $10/month (unlimited completions). Pro+: $39/month (1,500 premium requests). Business: $19/user/month. Enterprise: $39/user/month.

Custom pricing (~$60/user/month reported, 150 user minimum, 12-month contract). Estimated $108,000 annual minimum.

Free: 300K daily/1M monthly tokens. Pro: $25/month (10M tokens). Teams: $30/member/month (10M+ shared). Enterprise: Custom. 10% discount yearly.

Free: 3 videos/month (3 min, 720p). Creator: $29/month (unlimited 30 min, 1080p). Team: $39/seat/month (4K, 2 seats min). Enterprise: Custom pricing.

Free: 100 tasks/month. Professional: $19.99/month (750 tasks). Team: $69/month (2,000 tasks). Enterprise: Custom. Task-based pricing scales with usage.

Free: Limited use. Standard: $9.99/month (2,000 credits, ~20 videos). Pro: $29.99/month (~70 videos). Premium: TBA (~500 videos). Unlimited image gen until Jan 15, 2026.

Starter: Free (limited AI). Core: $20/month ($25 credits). Teams: $35/user/month ($40 credits). Enterprise: Custom. Agent: $0.25/checkpoint. Assistant: $0.05/edit.

Free: $5 credits/month. Premium: $20/month ($20 credits + $2 daily). Team: $30/user/month. Business: $100/user/month. Enterprise: Custom. Token-based: $0.50-$17.50 per 1M tokens.

Free: Limited AI (~50 uses). Pro: $15/month ($120 annual, ~500 AI uses). Teams: $10/user/month (min 3, pooled AI). Business: $16.67/month ($200 annual). Enterprise: Custom.

Free: $0. Pro: $12/month ($144 annual), $20/month (quarterly), $30/month (monthly). Enterprise: Custom pricing. 25% discount often available.

Creator: $49/month ($39 annual). Teams: $125/month ($99 annual). Business: Custom pricing. 7-day free trial. Unlimited words on all plans.

Chat: $29/month ($24 annual, 5 seats). Growth: $1,000/month (75 seats). Expansion: $2,000/month (150 seats). Scale: $3,000/month (200 seats). Enterprise: Custom.

Free: 10K credits/month. Starter: $5/month (30K credits). Creator: $11-$22/month (100K credits). Pro: $99/month (500K credits). Scale: $330/month (2M credits). Business: $1,320/month (11M credits). Enterprise: Custom.

Mistral's licensing strategy is nuanced. Mistral Large 2 is released under the Mistral Research License, which permits non-commercial use; commercial self-deployment requires a separate license. Its API pricing is $2.00 per million input tokens and $6.00 per million output tokens. In a significant move for the open-source community, Devstral 2 is released under a modified MIT license, making it fully open-source and permissively licensed for commercial use. During an initial free period, its AP

All the major DeepSeek models, including R1 and the V3.2 series, are released under the highly permissive MIT license, allowing for unrestricted commercial use. This has made them very popular among developers and startups. The pricing for API access reflects the models' efficiency. The standard DeepSeek-V3.2 model is priced at just $0.28 per million input tokens and $0.42 per million output tokens through DeepSeek's own API, with even lower costs for cached inputs. The sparse attention mechani

A major strength of the Qwen series is its licensing. All the mentioned models are released under the permissive Apache 2.0 license, which allows for broad commercial use without significant restrictions. This has fostered wide adoption. API access through providers like DeepInfra and Novita is very affordable. For instance, the Qwen3 32B model starts at $0.10 per million input tokens and $0.30 per million output tokens. The MoE variant, Qwen3 30B A3B, is priced similarly, highlighting the cost

All Grok-4 models are proprietary and accessed via the xAI API. The pricing structure clearly reflects the trade-off between intelligence and cost. The premium models, Grok-4 and Grok-4.1, are priced at $3.00 per million input tokens and $15.00 per million output tokens. The highly cost-efficient "Fast" variants, Grok 4 Fast and Grok-4.1 Fast, are priced at just $0.20 per million input tokens and $0.50 per million output tokens, making them exceptionally competitive for high-throughput automati

All Gemma 3 models are released with open weights under the Gemma license, which permits commercial use and distribution. For managed API access, providers like DeepInfra and Novita offer extremely competitive pricing. For example, the Gemma 3 27B model is priced at approximately $0.10 per million input tokens and $0.20 per million output tokens. The smaller 4B model is even cheaper, at $0.02 per million input tokens and $0.04 per million output tokens. This aggressive pricing makes the Gemma 3

The models are released under the Llama 4 Community License Agreement, which allows for broad use, modification, and distribution, including for commercial purposes, with some restrictions. As open-weight models, they can be self-hosted for free (excluding hardware costs). For those preferring managed access, a vibrant ecosystem of third-party API providers has emerged. Pricing is exceptionally competitive due to the models' computational efficiency. For Llama 4 Maverick, API input costs range

Claude Opus 4.8 is Anthropic's flagship large language model released May 28, 2026, achieving top rankings on SWE-bench Pro (69.2%) and the Artificial Analysis GDPval-AA leaderboard (1890 Elo). It introduces Dynamic Workflows for orchestrating hundreds of parallel subagents and Effort Control for adjustable reasoning depth, with a 1M token context window by default.

Gemini 3.1 Pro is Google DeepMind's flagship reasoning model released February 19, 2026, as a mid-cycle update to the Gemini 3 series. It achieves 77.1% on ARC-AGI-2 (more than double its predecessor), supports a 1-million-token context window, and introduces three-tier 'Thinking' levels for balancing speed and computational depth. It uses a Transformer-based Mixture-of-Experts architecture.

Gemma 4 is Google DeepMind's fourth generation of open-weight multimodal AI models, released March 31, 2026 under the Apache 2.0 license. The family includes four variants (E2B, E4B, 26B MoE, 31B Dense) optimized for deployment from edge devices to enterprise servers. The 31B Dense model ranks #3 among open models on Arena AI and achieves 89.2% on AIME 2026 mathematics benchmark.

Gemini 3.5 Flash is Google's high-efficiency multimodal language model released May 19, 2026 at Google I/O 2026. It features a 1M token context window and native multimodality (text, image, video, audio), designed to outperform previous Pro-tier models in coding and agentic tasks at approximately 40% lower cost than Gemini 3.1 Pro for those use cases.

ZAYA1-8B is a reasoning-focused mixture-of-experts (MoE) language model from Zyphra with 8.4 billion total parameters and only 760 million active parameters per token. It was trained entirely on AMD Instinct MI300X hardware and achieves competitive reasoning performance at a fraction of the compute cost of larger models.

Claude Sonnet 5 is Anthropic's most agentic Sonnet-class model, released June 30, 2026. It can make plans, use tools like browsers and terminals, and run autonomously at a level previously requiring larger and more expensive models. Performance is close to Claude Opus 4.8 at lower prices.

Gemini 3.8 Flash is Google's current Flash-tier language model, released September 2, 2026. Positioned for agentic workflows, coding, and complex reasoning tasks, it supersedes Gemini 3.7 Flash. A restricted-access cybersecurity variant, Gemini 3.8 Flash Cyber, is available through Google's Fairwind Program for trusted defenders.

Tencent Hy4 is a large-scale open-source foundation model released in preview on August 28, 2026. It features 770 billion total parameters with 49 billion active parameters in a Mixture-of-Experts architecture, supporting context windows exceeding 1 million tokens. Licensed under Apache 2.0, it targets coding, research, and productivity tasks.

Kimi K3 is a large-scale open-weight Mixture-of-Experts language model from Moonshot AI, with 2.8 trillion total parameters and 104 billion active parameters. API access launched July 16, 2026, with full weights released July 27, 2026. It supports a 1 million token context window and targets institution-scale deployments. Released under a bespoke Kimi K3 License with commercial conditions for large operators.

Meta Muse Spark 1.1 is a multimodal and agentic foundation model from Meta AI, released July 9, 2026 with API access. It is a separate model family from Llama 4, featuring a 1 million token context window and designed for multimodal reasoning and agentic workflows. Available through the Meta Model API.

Claude Opus 5 is Anthropic's flagship large language model, released July 24, 2026. It supersedes Claude Opus 4.8 as the most capable model in the Claude lineup, designed for complex reasoning, research, and agentic tasks. Available via the Anthropic API and Claude.ai.

GPT-6 Astra is OpenAI's frontier model released September 3, 2026, designed for high-level agentic tasks including software engineering, deep research, and complex computer-use workflows. It features a 1,050,000-token context window and multimodal input (text and image). Access was initially staged through a 'Daybreak Access' program before broader API and ChatGPT availability.

Gemini 3.7 Flash is Google's cost-focused model released August 13, 2026, optimized for agentic and multimodal workflows with configurable thinking levels. It supersedes Gemini 3.5 Flash and 3.6 Flash in the Flash tier, offering improved performance for coding and agentic tasks at competitive pricing.

Beam is Reflection AI's inaugural open-weight frontier model, released October 5, 2026. It is a sparse mixture-of-experts model with 501 billion total parameters and 23 billion active parameters per token, pretrained on 23.8 trillion tokens. Designed for coding, reasoning, and agentic workloads, Beam is positioned as a Western alternative to Chinese open-weight models with superior inference efficiency. Model weights, technical report, and developer artifacts are planned for Apache 2.0 release later in October 2026.

Claude Opus 5.5 is Anthropic's most capable model as of September 2026, released September 22, 2026. It is designed for long-running agentic coding and knowledge work, featuring a one-million-token context window and adaptive thinking. It includes strengthened cybersecurity, biology, and anti-distillation safeguards.

Claude Sonnet 5.5 is Anthropic's latest Sonnet-series language model, released September 28, 2026. It is positioned as a faster and lower-cost complement to Claude Opus 5.5, optimized for everyday professional tasks, bug fixing, and document creation. Pricing is $2 per million input tokens and $10 per million output tokens.

GPT-6.1 Sol is OpenAI's latest model in the GPT-6 family, released September 29, 2026. It is designed for coding, professional workflows, and computer use, and is available via the OpenAI API, ChatGPT Work, and Codex. OpenAI's Preparedness Framework classifies it as 'Critical' for cybersecurity capabilities.

Mistral Large 4 is a natively multimodal mixture-of-experts (MoE) frontier model with approximately 1.05 trillion total parameters and 52 billion active parameters per token. Released as a public API preview on October 6, 2026, it features a native 1.6 billion-parameter vision encoder and a 1-million-token context window. Trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's European data centers; open weights planned for release October 31, 2026.

Claude Haiku 5.5 is Anthropic's most capable, cost-efficient, and fastest small model, released October 7, 2026. It features a 1-million-token context window, 128K maximum output, and a two-tier pricing structure based on prompt length. The first Haiku model with adjustable thinking effort, it is designed for high-volume, latency-sensitive tasks such as classification, data extraction, and subagent workloads. Available via Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry.

Timestral is a 209-million-parameter probabilistic time-series forecasting model released by Mistral AI on October 15, 2026. It is a specialist model distinct from Mistral's general-purpose LLMs, designed for structured temporal prediction tasks rather than natural language generation.

Gemini 3.8 Flash Cyber is a specialized variant of Google's Gemini 3.8 Flash model, optimized for cybersecurity tasks including threat analysis, vulnerability assessment, and security code review. Released alongside Gemini 3.8 Flash in September 2026.

Mistral Large 4 is Mistral AI's current flagship large language model, released in October 2026. It succeeds Mistral Large 2 and offers improved reasoning, coding, and multilingual capabilities, available via the Mistral API and La Plateforme.

Kimi K3.1 Pro is a frontier language model from Moonshot AI, the first in the K3.1 series, with hosted availability from October 23-24, 2026. It is designed for coding, reasoning, and general-purpose tasks, with open weights announced for release on November 4, 2026.

GPT-6.1 Terra is a specialized model in OpenAI's GPT-6 family, released October 26, 2026, optimized for visual reasoning, computer use, coding, and automated evaluation tasks. Available via the API with identifier gpt-6.1-terra, and through selected ChatGPT Work and Codex surfaces.

DeepSeek V4.1 Flash is the current-generation model from DeepSeek, released September 10, 2026. It succeeds the V3.2 and R1 generations, representing DeepSeek's latest advancement in the V4 model family with improved performance and efficiency.

Meta Muse Wonder is a frontier foundation model released October 12, 2026, featuring Active Memory for retaining context across long sessions and Active Focus for persisting through difficult multi-step tasks. It is available through the Meta Model API and Muse Code platform, with a more capable 'Full' variant available under narrower subscription access.

MiniMax M3.1 Flash is an updated AI model from Chinese AI company MiniMax, launched October 19, 2026. It is a successor to the M3 model (placed on EU market June 1, 2026) with claimed major improvements in performance and efficiency, targeting enterprise and developer use cases.

Mistral Large 4 is Mistral AI's latest flagship model, with weights released October 29, 2026. Announced in early October 2026, it is positioned as competitive with leading Chinese models in certain areas per Mistral's CEO. License and exact API identifiers require first-party confirmation.

Reflection Beam 1.0 is an open-source sparse mixture-of-experts language model released October 16, 2026, with 501.3 billion total parameters and 23.4 billion active parameters per token. It features a 1-million-token context window, native tool-calling grammar, configurable reasoning effort, and an Apache 2.0 license. The BF16 checkpoint is approximately 1 terabyte.