ChatGPT
Broad all-purpose workStrongest default if you want one assistant across research, writing, image tools, and agent-style workflows.
Source audit: September 28, 2026 (OpenAI, Anthropic, DeepSeek API prices and ChatGPT plan prices) · August 11, 2026 (other details)
Compare AI Models: Hosted, Open Source, and Open Weight
Compare current hosted and open-model families with official sources, license context, standard API pricing where published, local deployment reality, and workflow guidance. Rankings are editorial judgments, not benchmark claims.
New open-model library
Eleven reviewed open-model families, with the details people actually search for.
Each page separates model access from ownership, open source from open weight, API price from infrastructure cost, and impressive context length from practical deployment. Start with the job you need the model to do, then open the detailed source record.
What is included
11 model pages9 providers covered1M context models explainedPrimary sources only
Provider
Open-weight
Kimi K3
Moonshot AI
Long-horizon coding agents
Kimi K3 License1M tokens
Open-weight
MiniMax M3
MiniMax
Agentic coding
Open-weight release, verify model-card terms1M tokens
Open source
Qwen3.6
Qwen / Alibaba
Local coding assistance
Apache 2.0Model and runtime dependent
Open source
GLM-5.2
Z.ai
Long-horizon coding agents
MIT1M tokens
Hosted API with published model ecosystem
DeepSeek V4
DeepSeek
Cost-sensitive API workflows
Verify current weight terms1M tokens
Open source
Mistral Small 4
Mistral AI
Multimodal document analysis
Apache 2.0256K tokens
Open source
Mistral Large 3
Mistral AI
Enterprise custom models
Apache 2.0Verify selected endpoint or checkpoint
Open model family
Gemma 4
Right-sized open-model experiments
Gemma 4 licenseModel variant dependent
Open-weight
gpt-oss-120b
OpenAI
Private-cloud reasoning services
Apache 2.0128K tokens
Open-weight
gpt-oss-20b
OpenAI
Local text reasoning
Apache 2.0128K tokens
Open-weight
Nemotron 3 Super
NVIDIA
Enterprise agent systems
NVIDIA Open Model LicenseLong-context agentic reasoning
Open model data
Reuse the source-dated comparison
Download the current model, pricing, workflow-fit, and official-source fields. Licensed under CC BY 4.0 with attribution to T-Minus AI.
Best overall default
ChatGPT
Best if you want one assistant that can cover writing, research, image work, and agent-style workflows.
Best for writing
Claude
Best for long-form writing, nuanced tone, and careful document-heavy work.
Best for Google users
Gemini
Best if your work already runs through Gmail, Docs, Drive, Meet, and the wider Google stack.
Best for live social context
Grok
Best when X-native signals, trending context, and real-time commentary matter.
Best low-cost API option
DeepSeek
Best for teams optimizing hard around cost-per-token and API efficiency.
Best open-weight option
Llama
Best for builders who want self-hosting, open weights, and more infrastructure control.
Fast answer
ChatGPT vs Claude vs Gemini vs Grok vs DeepSeek vs Llama
If you only want the short version: ChatGPT is the best all-purpose default, Claude is the best writing model, Gemini is the strongest fit for Google-native work, Grok is best for live social context, DeepSeek is one of the best cost-efficient API choices, and Llama is the best-known open-weight family for teams that want more control.
Claude
Writing and document workThe cleaner fit for long-form writing, document analysis, and careful structured output.
Gemini
Google-native workflowsA strong fit if your work happens inside Gmail, Docs, Sheets, Meet, and the Google ecosystem.
Grok
Live social and trend contextMost relevant when real-time X context and social signal tracking are central to the workflow.
DeepSeek
Low-cost API deploymentOne of the most cost-efficient ways to access capable models for coding and budget-sensitive developer workflows.
Llama
Open-weight controlThe best-known open-weight family when you want self-hosting, privacy control, or infrastructure flexibility.
Subscription decisions
When the model question becomes a buying decision
ChatGPT Plus vs Claude Pro
Broad OpenAI workbench vs Claude document depth.
Open decision pageGoogle AI Pro vs ChatGPT Plus
Standalone assistant vs Google-native bundle.
Open decision pageClaude Pro vs Google AI Pro
Writing and analysis vs Google workflow leverage.
Open decision pagePerplexity Pro vs ChatGPT Plus
Research-first subscription vs execution workbench.
Open decision pageSide-by-side specs
Compare AI Models, head-to-head.
| Model | Input /1M | Output /1M | Access | Best For |
|---|---|---|---|---|
| GPT-6 AstraOpenAI | $10 | $50 | ChatGPT Work and Codex (desktop app) | OpenAI's most capable model (1.05M context) |
| GPT-6 SolOpenAI | $2 | $10 | ChatGPT Plus, Pro, Business from Oct 14 | Complex coding and agentic work |
| GPT-6 LunaOpenAI | $0.10 | $0.50 | ChatGPT Free and Go from Oct 14 | Efficient, high-volume API work |
| GPT-5.6 SolOpenAI | $4 | $20 | OpenAI API | Previous generation: existing GPT-5.6 integrations (price cut August 21, 2026) |
| Claude Opus 5.5Anthropic | $4 | $20 | Claude Pro, Max, Team, Enterprise | Anthropic's recommended starting model for most workloads (1M context) |
| Claude Fable 5.1Anthropic | $10 | $50 | Claude API | Demanding reasoning and long-horizon agentic work |
| Claude Sonnet 5Anthropic | $2 | $10 | Claude | Balanced coding, agents, and document-heavy work |
| Claude Haiku 4.5Anthropic | $1 | $5 | Claude API | Low-cost, high-volume Claude API work (200K context) |
| Claude Opus 5Anthropic | $5 | $25 | Claude API (legacy) | Previous generation: existing Opus 5 integrations |
| Gemini 3.6 FlashGoogle | $1.50 | $7.50 | Gemini | Multimodal and agentic API work |
| Gemini 3.5 Flash-LiteGoogle | $0.30 | $2.50 | Gemini | High-volume, low-cost API tasks |
| Grok 4.5xAI | $2 | $6 | Grok | General work and API experimentation |
| Kimi K3Moonshot AI | ¥20 | ¥100 | Kimi API and certified inference partners | Long-horizon coding agents |
| MiniMax M3MiniMax | $0.30 | $1.20 | MiniMax API and Token Plan | Agentic coding |
| Qwen3.6Qwen / Alibaba | Check provider | Check provider | Qwen Studio and Alibaba Cloud Model Studio | Local coding assistance |
| GLM-5.2Z.ai | Model-specific price not listed | Model-specific price not listed | Z.ai general API and GLM Coding Plan | Long-horizon coding agents |
| DeepSeek V4DeepSeek | $0.15 to $0.30 | $0.60 to $1.20 | DeepSeek API | Cost-sensitive API workflows |
| Mistral Small 4Mistral AI | $0.15 | $0.60 | Mistral API and Mistral AI Studio | Multimodal document analysis |
| Mistral Large 3Mistral AI | Check provider | Check provider | Mistral AI Studio and selected cloud providers | Enterprise custom models |
| Gemma 4Google | No universal Gemma API rate | No universal Gemma API rate | Downloadable model family and Google AI ecosystem | Right-sized open-model experiments |
| gpt-oss-120bOpenAI | No OpenAI API price | No OpenAI API price | Self-hosted or third-party hosting providers | Private-cloud reasoning services |
| gpt-oss-20bOpenAI | No OpenAI API price | No OpenAI API price | Self-hosted or third-party hosting providers | Local text reasoning |
| Nemotron 3 SuperNVIDIA | Check provider | Check provider | Downloadable weights, NVIDIA services, and selected hosting providers | Enterprise agent systems |
Standard API costs per 1M tokens where the provider publishes a comparable USD rate. Open-model rows link to detail pages that separate API pricing from local operating cost. Latest source audit September 28, 2026; each row keeps its own verification date in the downloadable dataset.
OpenAI and Anthropic rows, and DeepSeek prices, were verified on September 28, 2026. Google, xAI, and the other open-model rows keep their August 11, 2026 audit date.
Claude Sonnet 5 costs $2 input and $10 output per million tokens. Anthropic's pricing page lists this as the standard price; the previously announced increase will not occur.
OpenAI says GPT-5.5 retires from ChatGPT, ChatGPT Work, and Codex on October 14, 2026. GPT-6 Sol replaces it on Plus, Pro, Business, Enterprise, and Edu; GPT-6 Luna replaces it on Free and Go.
Anthropic lists Claude Haiku 4.5 retirement as not sooner than October 15, 2026.
DeepSeek charges twice the off-peak rate during peak hours (01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday). The table shows the off-peak to peak range; the dataset records the off-peak rate.
Model access in consumer chat plans varies by provider, region, and plan. Check the linked provider page before purchasing.
AI Model Finder
Find Your Perfect AI Model
Answer a few quick questions and we'll recommend the best AI model for your needs.
Question 1 of 333% Complete
What will you primarily use AI for?
Quick Reference Guide
Which AI for Which Task?
Not sure which model to use? This matrix maps common use cases to the best AI models with clear reasoning.
| Use Case | Best Model | Why? | Alternative |
|---|---|---|---|
| Writing long-form content | Claude Sonnet 5 | A practical current Claude option for document-heavy drafting and revision workflows. | GPT-6 Sol |
| Quick research with sources | Perplexity Pro | Fast web discovery with citations and strong source traceability. | Gemini 3.6 Flash |
| Advanced reasoning & math | GPT-6 Astra | OpenAI lists Astra as its most capable model. At $10 / $50 per 1M API tokens, reserve it for the hardest problems. | Claude Fable 5.1 |
| Code generation & debugging | Claude Opus 5.5 | Anthropic's recommended starting model for most workloads, at $4 / $20 per 1M API tokens. | GPT-6 Sol |
| Excel/Word automation | Microsoft 365 Copilot | Native integration in Word, Excel, Outlook, and Teams workflows. | ChatGPT |
| Real-time social signals | Grok 4.5 | Strong fit when live X context is central to the workflow. | Perplexity Pro |
| Cost-effective API usage | GPT-6 Luna | OpenAI lists Luna for efficient, high-volume work at $0.10 input and $0.50 output per 1M tokens. | DeepSeek-V4.1-Flash |
| Privacy-focused browsing | Brave Leo | Browser-native assistant with privacy-first defaults and BYOM flexibility. | Local LLMs |
| Multi-language translation | Gemini 3.6 Flash | Strong multilingual handling with multimodal capabilities. | GPT-6 Sol |
| Technical documentation | Claude Sonnet 5 | A current Claude option for structured technical documentation workflows. | GPT-6 Sol |
| Brainstorming sessions | ChatGPT Plus | A broad consumer app for quick iteration and idea generation. | Gemini 3.6 Flash |
| Image-generation briefs | ChatGPT (image tools) | Strong prompt iteration and multimodal planning for visual direction. | Gemini 3.6 Flash |
Writing long-form content
CreatorsBest:Claude Sonnet 5
A practical current Claude option for document-heavy drafting and revision workflows.
Alt:GPT-6 Sol
Quick research with sources
ResearchBest:Perplexity Pro
Fast web discovery with citations and strong source traceability.
Alt:Gemini 3.6 Flash
Advanced reasoning & math
DevelopersBest:GPT-6 Astra
OpenAI lists Astra as its most capable model. At $10 / $50 per 1M API tokens, reserve it for the hardest problems.
Alt:Claude Fable 5.1
Code generation & debugging
DevelopersBest:Claude Opus 5.5
Anthropic's recommended starting model for most workloads, at $4 / $20 per 1M API tokens.
Alt:GPT-6 Sol
Excel/Word automation
CorporateBest:Microsoft 365 Copilot
Native integration in Word, Excel, Outlook, and Teams workflows.
Alt:ChatGPT
Real-time social signals
ResearchBest:Grok 4.5
Strong fit when live X context is central to the workflow.
Alt:Perplexity Pro
Cost-effective API usage
DevelopersBest:GPT-6 Luna
OpenAI lists Luna for efficient, high-volume work at $0.10 input and $0.50 output per 1M tokens.
Alt:DeepSeek-V4.1-Flash
Privacy-focused browsing
CorporateBest:Brave Leo
Browser-native assistant with privacy-first defaults and BYOM flexibility.
Alt:Local LLMs
Multi-language translation
CorporateBest:Gemini 3.6 Flash
Strong multilingual handling with multimodal capabilities.
Alt:GPT-6 Sol
Technical documentation
DevelopersBest:Claude Sonnet 5
A current Claude option for structured technical documentation workflows.
Alt:GPT-6 Sol
Brainstorming sessions
CreatorsBest:ChatGPT Plus
A broad consumer app for quick iteration and idea generation.
Alt:Gemini 3.6 Flash
Image-generation briefs
CreatorsBest:ChatGPT (image tools)
Strong prompt iteration and multimodal planning for visual direction.
Alt:Gemini 3.6 Flash
Picks are editorial judgments, not benchmark results. OpenAI, Anthropic, and DeepSeek entries were checked against official documentation on September 28, 2026; other entries date from the August 11, 2026 audit.
1
ChatGPT
OpenAI
Position Most adopted all-purpose assistant
Models GPT-6 Astra · GPT-6 Sol · GPT-6 Luna
"OpenAI lists GPT-6 Astra as its most capable model, GPT-6 Sol for complex coding and agentic work, and GPT-6 Luna for efficient, high-volume work. In ChatGPT, Astra is available in Work and Codex in the desktop app."
Official source 2
Microsoft Copilot
Microsoft
Position Best native fit for Microsoft 365 workflows
Models Workspace-centric Copilot experiences across Microsoft 365
"Deep integration with Outlook, Word, Excel, Teams, and enterprise controls. Best fit when your AI layer must sit inside existing Microsoft workflows."
Official source 3
Google Gemini
Google
Position Best multimodal coverage + price-performance range
Models Gemini 3.6 Flash · Gemini 3.5 Flash · Gemini 3.5 Flash-Lite
"Google lists Gemini 3.6 Flash as its most balanced model for speed and intelligence, with 3.5 Flash and Flash-Lite covering sustained agentic and lower-cost workloads."
Official source 4
Perplexity
Perplexity
Position Fastest research-first assistant for source-backed discovery
Models Standard · Pro · Max · Enterprise Pro
"Perplexity's official plan guide makes the product easier to slot into a stack: free discovery at Standard, deeper research and advanced model access on Pro, and higher-end throughput on Max and Enterprise tiers."
Official source 5
Claude
Anthropic
Position High-trust long-context and agent workflows
Models Claude Opus 5.5 · Claude Fable 5.1 · Claude Sonnet 5 · Claude Haiku 4.5
"Anthropic recommends Claude Opus 5.5 ($4 / $20 per 1M tokens, released September 22, 2026) as the starting model for most workloads. Fable 5.1 ($10 / $50) is for demanding reasoning and long-horizon agentic work, Sonnet 5 ($2 / $10) is the lower-cost balanced option, and Haiku 4.5 ($1 / $5) is the cheapest tier."
Official source 6
Grok
xAI
Position Current xAI model and API offering
Models Grok 4.5
"xAI lists Grok 4.5 at $2 per million input tokens and $6 per million output tokens in its official release update."
Official source 7
DeepSeek
DeepSeek
Position Low-cost API with off-peak pricing
Models DeepSeek-V4.1-Flash · DeepSeek-V4-Pro
"DeepSeek's API lists V4.1-Flash and V4-Pro, both with a 1M-token context window, and charges twice the off-peak rate during peak hours (01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday). The draw is cost pressure for API-heavy coding, reasoning, and builder workflows."
Official source 8
Brave Leo
Brave
Position Browser-native assistant with privacy focus
Models Model menu varies by free, premium, and BYOM options
"Built into Brave with privacy-centric defaults and model flexibility."
Official source 9
Meta Llama
Meta
Position Best open-source multimodal AI — free to self-host
Models Llama 4 Scout (17B) · Llama 4 Maverick (17B MoE) · Behemoth (training)
"Llama 4 Maverick beats GPT-4o and Gemini 2.0 Flash on major benchmarks while being fully open-weight. Multimodal (text + image), multilingual (12 languages), runs on a single H100."
Official source 10
Mistral AI
Mistral
Position European open-weight frontier AI with enterprise focus
Models Mistral 3 · Mistral Large 3 · Devstral 2
"Mistral 3 gives builders another serious open multimodal and multilingual model family, especially useful for privacy, EU data-residency, and on-premise deployment decisions."
Official source Pricing Guide
AI Model Pricing Comparison
Compare pricing across major AI models. Most offer free tiers with paid upgrades for power users.
| Model | Free Tier | Paid Tier | Price | Best For |
|---|---|---|---|---|
| ChatGPT | ✓ Free tier OpenAI moves Free and Go to GPT-6 Luna when GPT-5.5 retires from ChatGPT on October 14, 2026. Go costs $8/month. | Plus GPT-6 Sol replaces GPT-5.5 on Plus on October 14, 2026, alongside projects, tasks, custom GPTs, and higher limits across product features. Pro: from $100/mo 5x or 20x usage options. GPT-6 Sol replaces GPT-5.5 on Pro on October 14, 2026. | $20/mo | General execution and teams that want the broadest ChatGPT product surface |
| Claude | ✓ Free tier Core Claude access with memory across conversations, web search, and limited advanced features | Pro Higher limits with stronger access to research, code execution, projects, and Claude's premium feature surface Max 5×: $100/mo · Max 20×: $200/mo Heavy-usage access for professionals who hit Pro limits regularly | $20/mo | Long-form writing, technical docs, coding, and agent workflows |
| Google Gemini | ✓ Free tier Gemini app free tier with standard model access | Google AI Plus / Pro / Ultra More Gemini access and plan benefits vary by country and plan. Check the official plan page before buying. Google AI Ultra: region-dependent Highest limits and early access to advanced Gemini capabilities | Region-dependent | Google ecosystem power users, high-volume API workloads, multimodal tasks |
| Perplexity | ✓ Free tier Practically unlimited basic searches with very limited Pro searches and basic file uploads | Pro Extended Pro Search, advanced model access, image and video generation, and higher upload limits Max: $200/mo Highest individual tier for users who need more throughput than Pro | $20/mo | Source-backed research, market scanning, and multi-step agentic work |
| Microsoft Copilot | ✓ Free tier Basic Copilot chat and web-grounded responses | Copilot plans Microsoft 365 app integration and work-grounded productivity features Enterprise and agent add-ons available Advanced controls, governance, and agent capabilities | From $20/mo consumer; business plans vary | Office-heavy organizations and enterprise rollout on Microsoft stack |
ChatGPT
$20/mo✓ Free tier available
OpenAI moves Free and Go to GPT-6 Luna when GPT-5.5 retires from ChatGPT on October 14, 2026. Go costs $8/month.
Paid Tier: Plus
GPT-6 Sol replaces GPT-5.5 on Plus on October 14, 2026, alongside projects, tasks, custom GPTs, and higher limits across product features.
Pro: from $100/mo
5x or 20x usage options. GPT-6 Sol replaces GPT-5.5 on Pro on October 14, 2026.
Best for:
General execution and teams that want the broadest ChatGPT product surface
Claude
$20/mo✓ Free tier available
Core Claude access with memory across conversations, web search, and limited advanced features
Paid Tier: Pro
Higher limits with stronger access to research, code execution, projects, and Claude's premium feature surface
Max 5×: $100/mo · Max 20×: $200/mo
Heavy-usage access for professionals who hit Pro limits regularly
Best for:
Long-form writing, technical docs, coding, and agent workflows
Google Gemini
Region-dependent✓ Free tier available
Gemini app free tier with standard model access
Paid Tier: Google AI Plus / Pro / Ultra
More Gemini access and plan benefits vary by country and plan. Check the official plan page before buying.
Google AI Ultra: region-dependent
Highest limits and early access to advanced Gemini capabilities
Best for:
Google ecosystem power users, high-volume API workloads, multimodal tasks
Perplexity
$20/mo✓ Free tier available
Practically unlimited basic searches with very limited Pro searches and basic file uploads
Paid Tier: Pro
Extended Pro Search, advanced model access, image and video generation, and higher upload limits
Max: $200/mo
Highest individual tier for users who need more throughput than Pro
Best for:
Source-backed research, market scanning, and multi-step agentic work
Microsoft Copilot
From $20/mo consumer; business plans vary✓ Free tier available
Basic Copilot chat and web-grounded responses
Paid Tier: Copilot plans
Microsoft 365 app integration and work-grounded productivity features
Enterprise and agent add-ons available
Advanced controls, governance, and agent capabilities
Best for:
Office-heavy organizations and enterprise rollout on Microsoft stack
Pricing snapshot: ChatGPT plan prices and DeepSeek API prices verified on September 28, 2026; other rows date from the August 11, 2026 audit. Plan names, limits, and regional prices can change quickly.
Live references
Check benchmark dashboards and official provider updates
Artificial AnalysisLMArenaOpen LLM Leaderboard (Hugging Face)OpenAI ModelsClaude Models OverviewGoogle AI PlansxAI Models & PricingDeepSeek Docs
Links open in a new tab. OpenAI, Anthropic, and DeepSeek sources checked on September 28, 2026; other links date from the August 11, 2026 audit.
Blog Post
March 2026 model snapshot
Historical context. Check this page for the current source audit.
Free Tool
AI Plan Comparator — All Tiers, All Providers
ChatGPT, Claude, Gemini & Perplexity pricing compared
Frequently Asked Questions
AI Models — Common Questions
What is the best AI model in 2026?
There is no single best model for every workflow. OpenAI, Anthropic, and DeepSeek API prices and ChatGPT plan prices on this page were source-audited on September 28, 2026; other details date from August 11, 2026. ChatGPT is a broad all-purpose choice, Claude suits document-heavy work, Gemini fits Google-centric workflows, and each provider has different plan limits and API pricing.
What is the difference between ChatGPT and Claude in 2026?
ChatGPT is the broader product surface: more plan tiers, more built-in tools, deeper research and agent features, and stronger general-purpose coverage. Claude is the steadier choice for long documents, nuanced writing, Projects, Research, and focused document-heavy workflows.
How do Grok, DeepSeek, and Llama compare with ChatGPT and Claude?
Grok matters most when live X context is central to the workflow, DeepSeek is attractive when low API cost is the deciding factor, and Llama is the clearest open-weight option for teams that want self-hosting and infrastructure control. ChatGPT and Claude remain the strongest defaults for most general-purpose business and writing workflows.
Which AI model is best for coding in 2026?
For coding, the safest recommendation is to separate chat models from coding products. Claude remains a strong coding assistant, ChatGPT is useful for broad debugging and tool-heavy workflows, and GitHub Copilot is the best fit if you want coding help directly inside your IDE with team controls.
What is the best free AI model in 2026?
The best free AI model depends on the job. ChatGPT Free is the broadest starting point, Claude Free is strong for writing and documents, Gemini Free works well for Google-native workflows, and Llama or DeepSeek can be better if you care more about open weights or low-cost experimentation than polished consumer apps.
Should I pay for ChatGPT, Claude, Google AI, or Perplexity first?
Most people should pay for ChatGPT first if they want one broad default assistant. Choose Claude first if long-form writing, analysis, and document-heavy work matter more. Choose Google AI first if your workflow already lives inside Google. Choose Perplexity first if research with citations is the real bottleneck.