Runware review
Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026
We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings. Partners with a live offer in your country come first, then every other brand by score.
Our verdict
Runware suits developers who want LLM access and GPU work billed in the same place, with per-second GPU billing and an OpenAI-compatible API on one account. GPU time starts at $1.60 an hour, which leaves it ranked 64 of 95 on price.
Key facts
- Best for
- One account for LLM API calls and GPU rendering
- GPU models
- RTX PRO 6000, L40s, H100, H200, B200, B300, GB300 NVL72↗
- GPU hour from
- $1.60↗
- Billing granularity
- Per second↗
- Free credits
- $2↗
- Locations
- 3↗
- Compliance certifications
- SOC 2, ISO 27001, GDPR↗
- Legal entity
- Runware Ltd↗
- Headquarters
- United Kingdom↗
- Founded
- 2023↗
- Support channels
- Email, Help-centre, Community↗
- Pricing model
- Hourly billing
- Official website
- runware.ai
LLM APIs
- Input price from
- $0.076↗
- Output price from
- $0.50↗
- Models offered
- DeepSeek-V4.1-Flash, DeepSeek-V4-Flash-0731, DeepSeek-V4-Pro, GLM-5.3-Flash, GLM-5.3, GLM-5.2, GLM-5.1, GLM-4.7, Kimi K3, Kimi K2.6, MiniMax M3, MiniMax M2.7, Gemma 4 31B, Jev, GPT-6 Sol, GPT-6 Astra, GPT-5.6 Luna, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.5, Claude Opus 5.5, Claude Opus 4.8, Claude Sonnet 4.6, Claude Fable 5, Gemini 3.8 Flash, Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash, Gemini 3.5 Flash-Lite, Shieldstral 1.0 3B↗
- Largest context window
- 1,000,000↗
- OpenAI-compatible API
- Yes↗
Facts checked September 26, 2026, 6 sources
Apps and profiles
Pros and cons
Pros
- The hourly plan bills per second, so short GPU jobs stop costing a whole hour.
- $2 in free credits covers a trial of both the GPU and the LLM side.
- One OpenAI-compatible API fronts DeepSeek V4.1 Flash, DeepSeek V4 Flash 0731 and DeepSeek V4 Pro, so existing clients need little rework.
- Certifications are listed as SOC 2, ISO 27001 and GDPR, which answers the standard security review questions.
Cons
- GPU time starts at $1.60 an hour, which ranks 64 of 95 on price.
- Support runs on email, help centre and community only, so a blocked production job waits on email or the community.
- Compute is listed across 3 locations, which narrows where you can pin data residency.
Who it suits
- Teams that want image generation and LLM calls on one account rather than two vendor contracts.
- Builders running short GPU jobs, where per second billing keeps idle cost down.
- Developers already on an OpenAI-compatible client who would rather change a base URL than rewrite requests.
Who should look elsewhere
- SberCloud, whose catalogue in B300, H100 and A100 is built around A100 and V100 parts, if those legacy accelerators matter to you.
- ComputePrices.com, if the wider hardware table it tracks, H100 SXM, A100 SXM and H200, suits you better than one vendor's list.
The company behind it
Runware Ltd is the legal entity behind the platform, a GB company founded in 2023. The product line is narrow: GPU capacity and an LLM API sold through one console rather than the broad catalogue of a general cloud provider. Support runs on email, help centre and community, so most questions reach the team by email, the help centre or the community. Check the named entity on your invoice if your finance team contracts by legal name.
How Runware compares
Runware vs RunPod8.8 vs 9.5
Runware vs ScalaHosting8.8 vs 8.3
Runware vs Contabo8.8 vs 9.2
Runware vs Kamatera8.8 vs 9.5
Runware vs Cloudways8.8 vs 9.2
Runware vs PlanetHoster8.8 vs 8.7
Runware vs Acronis8.8 vs 9.9
Runware vs Aruba8.8 vs 9.3
Runware vs Decodo8.8 vs 9.1
Runware vs Hostinger8.8 vs 9.1
Runware vs MarsProxies8.8 vs 9.0
Runware vs hide.me8.8 vs 8.7
Runware vs Proton Drive8.8 vs 8.3
Frequently asked questions
What does a GPU hour cost on Runware?+
GPU time starts at $1.60 an hour on the hourly plan. That entry price puts Runware at 64 of the 95 providers compared on this measure, so treat it as an opening rate rather than a low one. Larger accelerators move the total further.
Do I pay for a whole hour or only for what I use?+
Charges accrue per second, so a short inference run does not cost a whole hour of GPU time. The plan itself is hourly, which is the simplest way to read the spend. Short jobs and bursty rendering queues are where that granularity matters most.
Can I try Runware before paying?+
New accounts get $2 in free credits, enough to send a few requests and render a test image before you commit to a plan. Treat it as a way to check the API shape rather than a lasting free allowance, since the entry rates start as soon as the credits run out.
Which models can I call through the LLM API?+
The catalogue runs from DeepSeek and GLM variants through Kimi, MiniMax and Gemma to GPT, Claude and Gemini families, listed in DeepSeek V4.1 Flash, DeepSeek V4 Flash 0731 and DeepSeek V4 Pro. The endpoint is OpenAI-compatible, so an existing OpenAI client can usually point at it with a base URL change.
How large a context can the models handle?+
The largest context window listed is 1,000,000 tokens, which sits at the top of the range tracked across providers. That suits long documents and large codebases in one request. The ceiling is set per model, so check the specific model before designing around the maximum.
Where does Runware run and what certifications does it hold?+
The company is registered in the GB under Runware Ltd and lists 3 locations for its compute. Certifications are listed as SOC 2, ISO 27001 and GDPR, which covers the usual buyer questions about security review. Confirm where workloads land before you sign.
How do I get help if something breaks?+
Support runs on email, help centre and community, so documentation and the community carry much of the load. Plan your own escalation path if a production job is blocked, since no phone or live chat channel is listed. That matters more for always-on rendering than for prototyping.
What do LLM calls cost?+
Entry pricing starts at $0.076 for input and $0.50 for output, so long answers cost more per request than short ones. The gap between the two is wide enough to matter for summarisation and extraction work, where output volume dominates the bill.
Sources
- runware.ai checked September 26, 2026
Backs: Compliance certifications, Support channels
“SOC 2 and ISO 27001 certified, GDPR aligned, with EU and US data residency.”
- runware.ai/about checked September 26, 2026
Backs: Founded
“2023 Founded”
- runware.ai/llm-api checked September 26, 2026
Backs: Largest context window, Models offered, OpenAI-compatible API
“Flagship V4 model with a native 1M token context window”
- runware.ai/pricing checked September 26, 2026
Backs: Free credits, GPU hour from, GPU models, Input price from, Pricing model, Output price from
“New users who sign up with a business email receive $2 in free credits to explore the platform and test various models.”
- runware.ai/serverless checked September 26, 2026
Backs: Billing granularity
“Early Access list rates per GPU-hour, metered per second.”
- runware.ai/terms checked September 26, 2026
Backs: Headquarters, Legal entity, Locations
“a company registered in the United Kingdom at Dns House, 382 Kenton Road, Harrow, Greater London, England HA3 8DP”
Compiled from 6 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.