Ollama
by Ollama
Run open-weight LLMs locally with one command, then burst to Ollama's cloud on the same API
Ollama is an MIT-licensed runtime for downloading and running open-weight language models on a laptop, workstation or server, with an optional paid cloud for larger models. It is for developers and IT teams who want private, local model access behind a simple CLI and REST API, and a way to scale up without changing code.
Ollama is an open-source tool, written in Go and released under the MIT licence, that packages open-weight models such as gpt-oss, Qwen, Gemma, DeepSeek, Kimi, GLM and MiniMax so a developer can pull and run one with a single command on macOS, Windows, Linux or Docker. A local server exposes a REST API on port 11434, with official Python and JavaScript libraries, and the `ollama launch` command wires local or cloud models into coding agents including Claude Code, Codex, Copilot CLI and OpenCode. The GitHub repository has about 184,000 stars. In March 2026 Ollama moved Apple Silicon inference onto Apple's MLX in preview. Ollama's cloud lets the same API call larger models hosted mainly in the United States, billed per million tokens at each model's rate, with cheaper off-peak pricing; the company says prompt and response data is never logged or trained on and that partner GPU providers must keep zero data retention. On July 9, 2026 Ollama raised a $65 million Series B led by Theory Ventures, bringing total funding to $88 million, with Benchmark (which led the Series A), 8VC and Y Combinator among its investors. The company reports 8.9 million developers, more than 67,000 community integrations and use inside 85% of the Fortune 500. Founders are CEO Jeffrey Morgan and Michael Chiang. Ollama competes with llama.cpp, LM Studio and Jan for local inference, and with vLLM for multi-user serving.
An engineering or IT leader who needs developers to use open-weight models on local hardware for privacy-sensitive work, with a paid cloud fallback for models too large to run locally.
One CLI and one REST API for local and hosted open models, so prompts can stay on the device and code does not change when a team needs a bigger model.
At a Glance
- Category
- Infrastructure & Cloud
- Pricing
- Freemium, Subscription, Usage-based
- Target Market
- CTOs, Enterprise Developers, IT Leaders, Data Scientists
- Deployment
- Open-source, Self-hosted, Hybrid
- Customers
- 8.9 million developers; used within 85% of the Fortune 500 (company-reported, July 2026)
Key Features
- ✓One-command local models
`ollama run` downloads and starts an open-weight model on macOS, Windows, Linux or Docker without manual quantization or build steps.
- ✓Local REST API
A server on localhost:11434 with chat endpoints and official Python and JavaScript libraries lets apps call local models like a hosted API.
- ✓Ollama cloud models
Larger open models run on Ollama's hosted GPUs through the same API, billed per million tokens with lower off-peak rates.
- ✓Coding agent launcher
`ollama launch` connects Claude Code, Codex, Copilot CLI, Droid and OpenCode to local or cloud models in one step.
- ✓MLX engine on Apple Silicon
A preview released in March 2026 runs models on Apple's MLX framework to speed up inference on Mac hardware.
- ✓Enterprise controls
The Enterprise plan adds model access controls and cost budgets per user and per API key for centrally managed usage.
Capabilities
Use Cases
- •Private code assistant
Developers run a coding model locally and connect it to their agent of choice so proprietary source code never leaves the laptop.
- •Offline document processing
Teams run structured extraction over sensitive documents and images on local hardware, avoiding the need to send files to a third-party API.
- •Hybrid local and cloud inference
An app uses a small local model for low-latency or sensitive calls and switches to a hosted large model for hard tasks with the same API.
- •Model evaluation before procurement
Platform teams test several open-weight models on their own prompts locally before committing GPU budget to a production serving stack.
Ideal For
Best For
- ✓Developers prototyping against open-weight models on a laptop or workstation with no cloud account
- ✓Privacy-sensitive work in healthcare, finance or government where prompts must stay on company hardware
- ✓Pointing coding agents such as Claude Code, Codex or OpenCode at local or open models through `ollama launch`
- ✓Teams that want a single API spanning local models and larger hosted open models billed per token
Not Ideal For
- ✗High-concurrency production serving for many users; vLLM or SGLang handle batching and throughput better
- ✗Teams that need SOC 2 reports, SSO or data residency documentation today; none is published on the pricing page and Team is still early access
- ✗Users who want to share model files with other runtimes, since Ollama stores models in its own registry format and switching means re-downloading
Integrations
Deployment
Market & Ratings
8.9 million developers; used within 85% of the Fortune 500 (company-reported, July 2026)
Market Analysis
Pros
- ✓Easiest local setup of the major runtimes; HN commenters defending it call it 'about 1000x easier to use' than llama.cpp
- ✓MIT licence and roughly 184k GitHub stars with a large integration ecosystem (Open WebUI, LangChain, Continue and others)
- ✓Clear published pricing for cloud use, with a stated no-logging, no-training policy
- ✓Local use stays free and unlimited at every plan level
Cons
- ✗HN users report it runs noticeably slower than llama.cpp on the same model and hardware
- ✗Default context length was capped low, which commenters say made models look worse than they are until users raised it
- ✗Models are stored in a proprietary registry format and hashed blobs other tools cannot read, which commenters describe as lock-in
- ✗Long-running criticism over crediting llama.cpp, including a 2025 HN thread titled 'Ollama violating llama.cpp license for over a year'
- ✗Enterprise governance is thin so far: no SOC 2 or SSO is listed on the pricing page and the Team plan is early access
Pricing
Free
$0
- ✓Unlimited local model use
- ✓Starter cloud credits
- ✓1 concurrent cloud request
Pro
From $20/mo ($200/yr)
- ✓$60/mo of included cloud usage
- ✓3 concurrent cloud requests
- ✓Larger Pro models
Max
From $100/mo
- ✓$300/mo of included cloud usage
- ✓10 concurrent cloud requests
- ✓Early access to newest models
Team (early access)
From $500/mo
- ✓$1,000/mo shared usage
- ✓Unlimited users
- ✓Centralized billing and admin
- ✓Priority support
Enterprise
Contact for pricing
- ✓Volume pricing
- ✓Model access controls
- ✓Cost budgets per user and API key
- ✓Private Slack channel
- ✓Custom security questionnaires
Running models on your own hardware is free and unlimited. Cloud use is metered in tokens at each model's rate (gpt-oss:20b at $0.07 input and $0.30 output per million tokens; kimi-k3 at $3 and $15), cheaper outside 12:00 to 18:00 UTC on weekdays. Paid plans include monthly credits that do not roll over, and extra credits can be bought on any plan. Requests over the concurrency limit queue. Prices as listed on ollama.com/pricing on October 22, 2026.
Security & Compliance
Connect
Sources
This page was written from 6 sources, 5 on domains other than ollama.com.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Inferact Platform
Managed vLLM inference from the creators of vLLM, on your own GPUs or on Inferact Cloud
Perplexity SPACE
Open-source Firecracker microVM sandbox for AI agents, the Apache 2.0 runtime behind Perplexity Computer
Gimlet Cloud
Multi-silicon inference cloud that splits AI agent workloads across GPUs, CPUs and accelerators
DigitalOcean Managed Agents
Managed agent runtime with microVM sandboxes, 16,000+ governed tools and serverless inference, billed on active CPU