One API. Multiple models.
Text, image, video, audio and embeddings.
Model[Kimi K3]
The AI-native platform
for builders and agents
Europe’s leading AI platform.
Bring models, GPU compute, and AI agents together, from first idea to global scale.
Building at scale? Book a meetingA connected AI ecosystem
One platform. Three capabilities.
Agents that work together.
Connect agents, tools and workflows.
Enterprise knowledge base
Market insight → Agent kitResearch brief → plan
Compute that scales.
Schedule GPUs around your workload.
Job queuejob_01 Dispatch
NVIDIA H100Running
NVIDIA H200Ready
NVIDIA B200Ready
Inference NVIDIA H100
Model library
Explore models for text, code, video, and images.
Text & code
Selected open models
DeepSeek V4.1 Flash
Million-token context with a leaner memory footprint.
Kimi K3
Code from long briefs and visual references.
Qwen3.8 Flash Next
Efficient hybrid attention for long text and images.
GLM-5.3 Flash
Multimodal coding agents. Just 18B active parameters.
Curated models for coding agents
One API for your coding agents. Choose a model for the task, then keep building with the same integration.
Image models
Creative example
Brand visuals
Detailed briefs, product imagery and targeted edits.
Explore a creative brief alongside preset concept art. Replay the visual process; no model request is sent.
Model capabilities
Loading artworkBriefComposeReveal
For enterprise
Build what’s next.
We’ll help run it.
Connect models, agents, and compute. Give every team room to build while keeping the whole picture in view.
ModelsAgentsGPUs
Commercial plan
Your volume. Your enterprise rate.
Discuss enterprise VIP pricing around your model mix, expected volume and support needs.
Explore the latest modelsAccess, spending and operations in one place
RunMux / Enterprise workspaceInteractive preview
Alex MorganEngineering
Role changes appear in Activity & alerts.
The right access for every team.
Explore roles, approved models and environment access in one workspace.
Deployment architecture
Global reach.
Local control.
Route requests by region and policy. Distribute workloads across nodes, or run inference in your own environment.
RunMux / Inference network
Policy-aware global routing01 / 03
NAM / 012 nodes
North AmericaSAM / 022 nodes
South AmericaEUR / 032 nodes
EuropeAFR / 042 nodes
AfricaASI / 053 nodes
AsiaOCE / 062 nodes
OceaniaANT / 071 node
AntarcticaOne gatewayRunMux
Drag to explore the network
#0001NAM → EURPaused
Continuous routingA shorter path to your models.
Bring requests through a shared gateway. Route to an eligible region around latency, availability and your data policy.
01Receive request02Apply routing policy03Serve from a region
Market insight agent
See what’s changing.
Decide what comes next.
Track brand risks, content trends, product demand, and sales leads. Put the findings to work in reports, creative briefs, and your CRM.
Choose a use case
Business goal
Track risks across the brand portfolio and prepare a response.
Anonymized cases from the provider brief, not RunMux customer references.
Case replay
CollectAnalyzeDeliver
Provider case · AnonymizedConsumer goods company
01
Risk coverage
Product quality, service complaints, and marketing risks
02
Brand portfolio
Monitor the parent brand and its sub-brands
03
Reporting cadence
Daily and weekly reports, plus major-incident reviews
Workflow summaryReplaying
Replaying the case summaryCase summary replay
Companion agents
ByteDance Character-led conversations
More than answers.
A sense of connection.
Shape a character, carry the story forward, and make room for everyday conversation.
- Character & toneDesigned for character-based conversation and roleplay.
- Conversation contextSupply prior turns to help the story stay connected.
- Image-led storiesBring images into character conversations and storytelling.
Persistent memory needs app-managed storage and retrieval. Capabilities depend on the model version and integration.
NoriSeed-sc · Character preview
Found this quiet little corner. Can we stay here for a while?
NoriTyping a reply
Responding…
Scripted preview · Authored dialogue
Model evidence
The task comes before the score.
Companion and general capabilities use different evaluations. Explore each evidence set independently.
Seed-sc / Character 0628
See the difference.
Character 0628
+7.62 pp
vs Seed-2.0-lite
Rank 1 of 6 in this comparison.Your next chapter