Filtering by: llm · clear filter
Liquid AI's LFM2.5 24B A2B is now on NanoGPT.
A text reasoning model with 24B total and 2.45B active parameters. It thinks before answering, supports tool calling and structured output, and has a 128K-token context.
https://nano-gpt.com/conversation?model=liquid%2Flfm-2.5-24b-a2b
Gemini 4 Argon now also has Flex and Priority tiers on NanoGPT.
Flex is 50% cheaper than Standard when you can wait. Priority costs 80% more for higher-priority processing. Choose in model settings or set service_tier in the API.
https://nano-gpt.com/conversation?model=google%2Fgemini-4-argon
Gemini 4 Argon is now on NanoGPT.
Google's flagship reasoning model for coding, science and multi-step agents. Work with text, images, audio, video and PDFs, with tool calling, structured output and a 1M-token context.
https://nano-gpt.com/conversation?model=google%2Fgemini-4-argon
A new option for creative writing and roleplay: TheDrummer's Skyfall 31B v4.2.
Now on NanoGPT with a 16K context window, included in the subscription at the standard 1x rate.
https://nano-gpt.com/conversation?model=TheDrummer%2FSkyfall-31B-v4.2
Meta's Muse Wonder is live on NanoGPT.
Work with text, images, audio, video and PDFs in a 1M-token context. Reasoning, tool calling and structured output are supported.
GPT-6.1 Sol now supports Ultrafast on NanoGPT.
Select it in chat settings or use service_tier: "ultrafast" in API requests for faster token generation.
Ultrafast costs 6x Standard pricing.
https://nano-gpt.com/conversation?model=openai%2Fgpt-6.1-sol
Gemma 4 Chimera X Heretic and Scarlet are now on NanoGPT.
Two new finetunes for creative writing, dialogue and roleplay, with image understanding and 262K context.
Both are included in the subscription.
https://nano-gpt.com/conversation?model=gemma-4-26b-a4b-it-chimeraxheretic
https://nano-gpt.com/conversation?model=gemma-4-31b-it-scarlet
Auto Model now decides how much thinking each message needs.
Quick questions can skip reasoning; harder tasks get more. Prefer to set it yourself? Choose a fixed level with the slider, then switch Back to Auto anytime.
Mistral Large 4 Preview is now on NanoGPT.
Work with long documents, code and images in a 512K-token context. Supports reasoning, tool calls and structured output.
https://nano-gpt.com/conversation?model=mistralai%2Fmistral-large-4
Claude Haiku 5.5 now live on NanoGPT!
https://nano-gpt.com/conversation?model=anthropic%2Fclaude-haiku-5.5
Apex Flash 1 and Apex Flash 1 Abliterated are now on NanoGPT.
GLM 5.3 Flash cybersecurity finetunes for code investigation and tool use. Both included in the subscription.
Abliterated is experimental, with modified refusal behavior.
https://nano-gpt.com/conversation?model=z-ai%2Fglm-5.3-flash-apex
Eight new Qwen 3.8 27B finetunes are now on NanoGPT: Antislop, DarkIdol, Fimi, Kimiko, Kiwen, Novelist, Swift and Synthia.
All have 262K context and are included in the subscription at the standard 1x usage rate.
Native web search is now available on supported OpenAI, Claude, Gemini and Grok models in NanoGPT.
Choose Native in the Web menu. Models without native search support fall back to Linkup.
Ling 3.1 Flash is now on NanoGPT, included in the subscription.
inclusionAI's hybrid reasoning model supports coding, planning and tool use, with 256K context and up to 32K output tokens.
Thinking can be switched on or off.
https://nano-gpt.com/conversation?model=inclusionai%2Fling-3.1-flash
GPT-6.1 Sol Pro is now on NanoGPT.
The same Sol model with Pro reasoning for difficult coding, analysis and planning. 1.05M-token context, image understanding and tool calling.
https://nano-gpt.com/conversation?model=openai%2Fgpt-6.1-sol-pro
GPT-6.1 Sol is now available on NanoGPT.
Built for complex coding, computer use and professional work, with a 1.05M-token context, image understanding, reasoning and tool calling.
https://nano-gpt.com/conversation?model=openai%2Fgpt-6.1-sol
Claude Sonnet 5.5 and Thinking are on NanoGPT.
1M context, image understanding and tool use. Anthropic reports 70.6% on Terminal-Bench 4.0 (Sonnet 5: 10.3%) and 30%+ faster output than Sonnet 5.
https://nano-gpt.com/conversation?model=anthropic%2Fclaude-sonnet-5.5
https://nano-gpt.com/conversation?model=anthropic%2Fclaude-sonnet-5.5%3Athinking
CyberHeabsy is now on NanoGPT.
A lower-refusal text model for authorized security research, red-team practice, roleplay and creative writing. Optional reasoning, tool calling and a 1M-token context.
https://nano-gpt.com/conversation?model=heabsy%2Fcyberheabsy
LongCat 2.5 Preview is now on NanoGPT and included in the subscription.
Meituan's new model supports image understanding, coding, tool use and extended thinking, with a 1M-token context.
GLM 5.3 Uncensored is now on NanoGPT.
The full GLM 5.3 model, adapted for lower-refusal writing, coding, reasoning and tool use. 1M context.
Experimental: expect unusual behavior or inconsistent response quality.
https://nano-gpt.com/conversation?model=z-ai%2Fglm-5.3-uncensored
Perceptron Mk1.5 is now on NanoGPT.
Understand images, video and audio, track objects across frames, and use tools to act on what the model sees. Built for visual reasoning and physical agents.
https://nano-gpt.com/conversation?model=perceptron%2Fperceptron-mk1.5
Qwen3.8 27B Uncensored TEE is now on NanoGPT.
A lower-refusal finetune for writing, coding and image analysis, with reasoning, tools and 262K context. Inference runs inside a provider-side Trusted Execution Environment.
https://nano-gpt.com/conversation?model=TEE%2Fqwen3.8-27b-uncensored
MiMo V2.6 Flash Uncensored is now on NanoGPT.
A lower-refusal, text-only finetune with 1M context, reasoning and tool calling.
https://nano-gpt.com/conversation?model=xiaomi%2Fmimo-v2.6-flash-uncensored
Solar Mini 4 + Thinking are now on NanoGPT.
524K context; $0.05 input / $0.20 output per million tokens. PAYG only.
https://nano-gpt.com/conversation?model=upstage%2Fsolar-mini4
https://nano-gpt.com/conversation?model=upstage%2Fsolar-mini4%3Athinking
Ember 1 is on NanoGPT in research preview.
Fireworks' Kimi K3-based model: images, tools and 1M context.
Fireworks describes Ember as a research preview, initially offered for two weeks on its serverless platform. Continued availability depends on demand.
Lumen Stealth is revealed: meet Kitani's Clover 1 150B Preview on NanoGPT.
200K context, images, reasoning and tools. Included at regular 1x subscription input usage; the 0.1x promotion has ended.
Clover remains a preview while reliability improves, so occasional interruptions are still possible. Existing Lumen Stealth chats and API IDs still work.
https://nano-gpt.com/conversation?model=kitani%2Fclover-1-150b
Aion 3.5 and Aion 3.5 Mini are now on NanoGPT for roleplay and long-form storytelling, both with 262K context.
Aion 3.5:
https://nano-gpt.com/conversation?model=aion-labs%2Faion-3.5
Aion 3.5 Mini:
https://nano-gpt.com/conversation?model=aion-labs%2Faion-3.5-mini
GPT-6 Sol and Luna now support Flex and Priority on NanoGPT, including Pro variants.
Flex: lower cost when latency matters less. Priority: higher-priority processing.
API: set service_tier to "flex" or "priority".
GPT-6 Sol:
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-sol
GPT-6 Sol Pro:
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-sol-pro
GPT-6 Luna:
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-luna
GPT-6 Luna Pro:
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-luna-pro
Qwen3.8 Max Prime is now on NanoGPT.
A higher-throughput Max variant for coding and long-running agents, with 1M context, image/video input, reasoning and tools.
$4 input / $12 output per million tokens. PAYG only.
https://nano-gpt.com/conversation?model=qwen%2Fqwen3.8-max-prime
Lumen Stealth is on NanoGPT for a temporary evaluation.
200K context, vision and reasoning. Included in the subscription at 0.1x input usage.
Prompts and responses are logged and may be reviewed. Do not send sensitive or confidential information.
https://nano-gpt.com/conversation?model=nano%2Flumen-stealth
GPT-6 Sol and GPT-6 Luna are now on NanoGPT.
Sol for demanding coding, reasoning and agentic work; Luna for fast chat, classification and lightweight agent workflows.
Both support image input, tool calling and a 1.05M-token context. Pay-as-you-go, not included in the subscription.
Claude Opus 5.5 is now available on NanoGPT.
Anthropic's flagship model for agentic coding, computer use and long-running tasks, with a 1M-token context, image understanding, reasoning and tool calling.
Pay-as-you-go: $4 input / $20 output per million tokens. Not included in the subscription.
https://nano-gpt.com/conversation?model=anthropic%2Fclaude-opus-5.5
Nex N2.5 Mini + Pro are now on NanoGPT, both included in the subscription.
Open-weight coding models with vision and 262K context.
Pay-as-you-go per 1M tokens: Mini: $0.025 in / $0.10 out Pro: $0.075 in / $0.25 out
https://nano-gpt.com/conversation?model=nex-agi%2Fnex-n2.5-mini
https://nano-gpt.com/conversation?model=nex-agi%2Fnex-n2.5-pro
Lower prices for Abliterated models on NanoGPT.
Input per 1M tokens: Abliterated Model: $3 -> $1 Abliterated Model Large + Large V2: $5 -> $3
Cached input is cheaper too. Output prices are unchanged.
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model-large
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model-large-v2
Grok 4.7 is now on NanoGPT.
500K context, vision and tool calling for coding and long-running tasks.
From $1.60 input / $4.80 output per 1M tokens. Higher rates above 200K input.
Xiaomi MiMo V2.6 Flash and Pro are now available on NanoGPT and included in the subscription at 1x usage.
Both support text, images, video and audio, with a 1M-token context window, reasoning and tool calling for coding and agent workflows.
https://nano-gpt.com/conversation?model=xiaomi%2Fmimo-v2.6-flash
https://nano-gpt.com/conversation?model=xiaomi%2Fmimo-v2.6-pro
Writing a novel, building characters, or running a long-form roleplay?
Qwen 3.8 27B Hemingway is now on NanoGPT: 262K context, image understanding and configurable reasoning.
Included in the subscription at 1x usage.
https://nano-gpt.com/conversation?model=qwen%2Fqwen3.8-27b-hemmingway
Step 5 Preview is now on NanoGPT.
StepFun's new flagship brings a 1M-token context window, image and video understanding, tool calling and structured output for coding and agent workflows.
$1 input / $2.70 output per million tokens.
https://nano-gpt.com/conversation?model=stepfun%2Fstep-5-preview
Three new options for defensive security research:
Qwen 3.8 27B, Gemma 4 26B A4B and GLM 5.3 Flash Cybersecurity are now on NanoGPT.
Image understanding + tool calling. All included in the subscription at 1x usage.
These variants retain moderation for illegal activities.
https://nano-gpt.com/conversation?model=qwen%2Fqwen3.8-27b-cybersecurity
https://nano-gpt.com/conversation?model=google%2Fgemma-4-26b-a4b-it-cybersecurity
https://nano-gpt.com/conversation?model=z-ai%2Fglm-5.3-flash-cybersecurity
Pareto runs several AI models on your request and returns one selected or synthesized answer.
Now on NanoGPT: text + images, tool calling and 262K context.
Experimental. Its underlying model mix, behavior and performance may change.
Moderate the answer as well as the prompt.
NVIDIA Nemotron 3.5 Content Safety is now on NanoGPT: safe/unsafe classification for user messages and assistant responses, with optional reasoning.
https://nano-gpt.com/conversation?model=nvidia%2Fnemotron-3.5-content-safety
Ollama Cloud is now a provider option on NanoGPT.
Run supported models including Kimi K3, DeepSeek V4.1 Flash, GLM 5.3 Flash and Gemma 4 through Ollama.
Choose Ollama in the provider selector, or set X-Provider: ollama in API requests.
27B reasoning with ternary weights.
Ternary Bonsai 2 27B is now on NanoGPT, included in the subscription.
PrismML reports 98.2% of Qwen3.8-27B's aggregate benchmark performance. With vision, tool calling and a 262K context window.
https://nano-gpt.com/conversation?model=prism-ml%2Fternary-bonsai-2-27b
DeepSeek V4.1 Flash, running inside a Trusted Execution Environment.
Now on NanoGPT with a 1M-token context, text and image understanding, reasoning and tool calling.
https://nano-gpt.com/conversation?model=TEE%2Fdeepseek-v4.1-flash
Who said what, and what needs doing next?
Qwen3.8 Omni Flash is now on NanoGPT. Analyze audio, video, images and text for meeting summaries, transcripts and action items.
Nearly 1M context, reasoning and tool calling.
https://nano-gpt.com/conversation?model=qwen%2Fqwen3.8-omni-flash
DeepSeek V4.1 Flash and Thinking are now included in NanoGPT subscriptions.
Text, images and tool calling. Rate-limited beta, not for production. Provider may log data and use it for training. Don't send sensitive data.
https://nano-gpt.com/conversation?model=deepseek%2Fdeepseek-v4.1-flash
Fugu Max is now on NanoGPT.
Sakana AI's multi-agent system coordinates specialist models for coding, reasoning and tool use, with a focus on cost-performance.
Try it on a problem where your usual model gets stuck.
You can now manage NanoGPT model presets through the API.
Create, inspect, update and delete saved configurations with a model, instructions and generation or routing defaults. Call them with presets/. Explicit request settings override saved defaults.
Manage presets:
API specification:
Agnes 3.0 Flash is now on NanoGPT.
A low-cost model for coding, tool use and multi-turn agent tasks, with image understanding, optional thinking and a 512K context window.
Ling 3.0 Flash VL is now on NanoGPT and included in your subscription.
Analyze images, charts, documents and videos with reasoning and tool calling.
256K context, with 5.5B active parameters.
https://nano-gpt.com/conversation?model=inclusionai%2Fling-3.0-flash-vl
TheDrummer/Artemis v1.1 is now on NanoGPT, included in the subscription.
A Gemma 4 31B fine-tune for creative writing, expressive character dialogue and roleplay, with a 262K context window for longer stories.
https://nano-gpt.com/conversation?model=TheDrummer%2FArtemis-v1.1
Qwen3 8B TEE is now on NanoGPT.
A compact open-weight model for chat, coding, reasoning and tool use, running inside a Trusted Execution Environment with attestation support.
More GPT 6 Astra options on NanoGPT: Pro reasoning mode for harder tasks, Flex and Priority service tiers, plus Batch API support for Chat Completions and Responses workloads.
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-astra-pro
Synth 2.5 Flash and Pro Preview are now on NanoGPT, both included in the subscription.
Two low-cost models for role-play, character dialogue and collaborative storytelling.
Flash:
https://nano-gpt.com/conversation?model=synth-2.5-flash
Pro:
GPT 6 Astra is live on NanoGPT.
OpenAI calls Astra its most capable model for the hardest end-to-end work, including software engineering, computer use, browsing, science, and professional tasks.
Try it:
https://nano-gpt.com/conversation?model=openai%2Fgpt-6-astra
Abliterated Model Large V2 now supports native tool calling and structured output on NanoGPT.
The unrestricted GLM-5.3-based model also supports prompt caching and has a 1M-token context.
Try it:
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model-large-v2
Runware is now available as an opt-in NanoGPT provider.
Its self-hosted routes for GLM 5.2, GLM 5.3, GLM 5.3 Flash, DeepSeek V4 Flash, and GPT OSS 120B are classified as Zero Data Retention. They support streaming, tool calling, reasoning, and prompt caching.
Runware is currently serving GLM 5.3 at up to 500 tokens per second.
Choose Runware through provider selection or pass X-Provider: runware in API requests.
Qwen3.8 Max 0902 is now available on NanoGPT.
The new 2.4T-parameter checkpoint supports a 1M-token context, multimodal input, selectable thinking, tool calling, structured output, and stronger coding and agent workflows.
Try it:
https://nano-gpt.com/conversation?model=alibaba%2Fqwen3.8-max-0902
Claude Fable 5.1 is now available on NanoGPT.
It improves on Fable 5 across the board, with the biggest gains in agentic coding, long-running workflows, and knowledge work. It is especially strong at long code refactors, front-end and visual code generation, finance, and analysis, while producing more concise plans and summaries.
Try it:
https://nano-gpt.com/conversation?model=anthropic%2Fclaude-fable-5.1
Mercury 2.5 Preview is now available on NanoGPT and temporarily included in subscriptions.
Inception's latest diffusion reasoning model reaches up to 1,107 tokens/sec, with tunable reasoning, parallel tool calls, schema-aligned JSON output, and a 10+ point intelligence gain over Mercury 2.
Try it:
https://nano-gpt.com/conversation?model=inception%2Fmercury-2.5-preview
Abliterated Model Large V2 is now on NanoGPT.
This uncensored GLM-5.3-based model has a one-million-token context window for harder reasoning, evaluation workloads, long documents, and demanding conversations.
Try it:
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model-large-v2
GLM 5.3 is now available in a private TEE on NanoGPT.
Get confidential inference with attestation support for long-horizon coding, autonomous agent workflows, tool use, vulnerability research, and complex engineering.
Try it:
Kimi K3 pricing just dropped on NanoGPT.
Standard non-ZDR input, cached input, and output rates are now about one-third lower. This reduction does not apply to ZDR, Private Mode, or Batch API requests.
Try Kimi K3:
https://nano-gpt.com/conversation?model=moonshotai%2Fkimi-k3
Abliterated Model and Abliterated Model Large are now available on NanoGPT.
Abliterated Model supports text, images, structured output, and long-context work with a 262K context window. Abliterated Model Large is text-only, based on GLM-5.2, with a one-million-token context window for harder reasoning and long documents.
Try them:
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model
https://nano-gpt.com/conversation?model=abliteration-ai%2Fabliterated-model-large
Gemma 4 26B A4B Uncensored is now available on NanoGPT and included in the subscription at 1x included-input usage!
This open-weight multimodal mixture-of-experts model is tuned for fewer refusals across chat, coding, tool use, image understanding, and long-context work.
Try it:
https://nano-gpt.com/conversation?model=google%2Fgemma-4-26b-a4b-uncensored
Qwen 3.8 27B Uncensored is now available on NanoGPT and included in the subscription at 1x usage.
This open-weight FP8 model is tuned for fewer refusals across chat, coding, tool use, image understanding, and long-context work.
Try it:
DeepSeek V4 Flash Vision Exp is now available on NanoGPT and included in the subscription at 1x usage.
It adds image understanding while retaining reasoning, coding, tool use, agent capabilities, and a 1M-token context window.
Try it:
Ornith 1.5 35B and its Thinking variant are now available on NanoGPT and included in the subscription at 1x included-input usage.
Ornith reports 79 on SWE-bench Verified, 68.5 on Terminal-Bench 2.1, and 22 on DeepSWE, with about 3B active parameters.
Try it:
https://nano-gpt.com/conversation?model=ornith-ai%2FOrnith-1.5-35b-a3b
Qwen3.8 27B and its Thinking variant are now included in the NanoGPT subscription at 1x included-input usage.
Use regular for fast, direct answers or Thinking for deeper reasoning, coding, tools, and agent tasks.
Try it:
DeepSeek V4 Pro 0813 and its Thinking variant are now available in a Trusted Execution Environment on NanoGPT.
Get private coding, tool use, cybersecurity, and agent workflows with attestation support and a 1M-token context window.
Try it:
https://nano-gpt.com/conversation?model=TEE%2Fdeepseek-v4-pro-0813
Qwen3.8 27B and Qwen3.8 27B Thinking are now available on NanoGPT.
Qwen reports 73.0 on Terminal Bench 2.1, 61.7 on SWE-bench Pro, 79.0 on QwenSWEBench, and 90.3 on LiveCodeBench v6.
Try it:
DeepSeek pricing changes at 16:00 UTC today.
The Cheaper V4 Pro and V4 Flash 0731 variants will be retired. Existing IDs will keep working through the regular versions. V4 Pro and V4 Pro 0813 will count at 2x subscription usage.
Details:
DeepSeek pricing changes at 16:00 UTC today.
The Cheaper V4 Pro and V4 Flash 0731 variants will be retired. Existing IDs will keep working through the regular versions. V4 Pro and V4 Pro 0813 will count at 2x subscription usage.
Details:
Muse Glimmer 30B is now available in a Trusted Execution Environment on NanoGPT.
Run Meta's multimodal coding and agent model inside a TEE with attestation support.
Try it:
https://nano-gpt.com/conversation?model=TEE%2Fmuse-glimmer-30b
Nemotron 3.5 Lightning is live on NanoGPT. Nvidia reports 60% on SWE-bench Verified.
https://nano-gpt.com/conversation?model=nvidia%2Fnemotron-3.5-lightning
Nemotron 3.5 Lightning is live on NanoGPT. Nvidia reports 60% on SWE-bench Verified.
https://nano-gpt.com/conversation?model=nvidia%2Fnemotron-3.5-lightning
Nemotron 3.5 Lightning is live on NanoGPT and included in the subscription.
Nvidia reports 60.0% on SWE-bench Verified.
Try:
https://nano-gpt.com/conversation?model=nvidia%2Fnemotron-3.5-lightning
Gerra is now available as a provider on NanoGPT!
Run a broad range of open models with streaming, tool calling, reasoning, structured outputs, and prompt caching. Many models are served in FP8, and Gerra’s production inference API is zero-retention: prompts and outputs are processed in memory, never stored or logged, and never used for training.
Select Gerra in provider settings, add :gerra to API model names, or connect your own Gerra key through BYOK:
#AI #Privacy #LLM #API #OpenSource #NanoGPT
For open source models we now link directly to the model weights on the model's page.
Our goal with the model pages is simple: we want to have all the information you might want about a model on one page.
Model details, pricing, weights, benchmarks, all in one overview.
See GLM 5.2 for an example:
https://nano-gpt.com/models/text/zai-org/glm-5.2
#AI #Models #OpenSource #LLM #GLM #NanoGPT
The Batch API now supports more models for high-volume asynchronous jobs, including DeepSeek V4 Flash, DeepSeek V4 Pro, GLM 5.1, GLM 5.2, GPT OSS 120B, GPT OSS 20B, Kimi K2.5, Kimi K2.6, Kimi K2.7 Code, MiniMax M2.7, MiniMax M3, Nemotron 3 Ultra NVFP4, Qwen3.6 Plus, and Qwen3.7 Plus.
These are intended for developers running larger jobs that can wait for asynchronous completion. Batch jobs can take up to 24 hours.
#AI #API #Batch #Developers #LLM #Models #NanoGPT
Fugu Ultra is now available!
It coordinates a pool of expert agents to tackle hard problems where answer quality matters most.
Fugu Ultra scores 73.7 on SWE Bench Pro, 82.1 on TerminalBench 2.1, 93.2 on LiveCodeBench, 90.8 on LiveCodeBench Pro, and 50.0 on Humanity's Last Exam, putting it ahead of listed public frontier baselines on several coding, agentic, scientific, and reasoning tests.
Try it for yourself!
#AI #Agents #Coding #LLM #Reasoning #Fugu #NanoGPT
GLM 5.2 is now available on NanoGPT and included in the NanoGPT subscription at 2x included-input usage.
Built for coding, agentic workflows, long-context chat, and tool use. Use the regular version for faster direct answers, or the Thinking variant for harder reasoning-heavy work.
#AI #Agents #Coding #LLM #GLM #NanoGPT
