250+ models. 50+ providers.
Frontier intelligence. Open-weight freedom.
| Model family and provider | Relative spend | Spend and share |
|---|---|---|
| Kimiby Moonshot AI | €70026.5% of total spend | |
| Claudeby Anthropic | €60022.7% of total spend | |
| Z.aiby Z.ai | €36013.6% of total spend | |
| Qwenby Alibaba | €2609.8% of total spend | |
 DeepSeekby DeepSeek | €2409.1% of total spend | |
 Muse Sparkby Meta | €2007.6% of total spend | |
 Mistralby Mistral AI | €1606.1% of total spend | |
 OpenAIby OpenAI | €1204.5% of total spend |
Claude Code780
Cursor312
Open WebUI156Bring Openbridge to your coding tools, chat apps, automations, and your own software. One place to manage AI across them all.
CursorClaude Code
WindsurfOpen WebUI
GitHub Copilot
n8n
ClineWe’re building smart limits, approved fallbacks, and routing to the right model for each task. Premium AI when you need it. Lower costs when you don’t.
$697$10,000| Model | Input | Output |
|---|---|---|
| Sonnet 4.6 · baseline | $3.00 | $15.00 |
| Kimi K2.6 · 10% | $0.95 | $4.00 |
| Qwen3.5 9B · 30% | $0.17 | $0.25 |
| DeepSeek V4 Flash (0731) · 30% | $0.22 | $0.66 |
| GPT-OSS 120B · 30% | $0.15 | $0.60 |
The logos illustrate the wider model selection. The calculation uses the four open-weight models listed above and Sonnet 4.6 as the premium baseline.
Example token allocation: 10% Kimi, 30% Qwen, 30% DeepSeek, and 30% GPT-OSS. Kimi, DeepSeek, and GPT-OSS use Fireworks standard rates; Qwen uses Together rates.
Assumes four input tokens per output token, unchanged total token volume, and the same token mix for every model. A $10,000 all-Sonnet baseline becomes $696.67 in model API costs. All work must be suitable for this mix; capabilities differ.
Published prices checked 7 September 2026. No caching or batch discounts. Excludes Openbridge markup, other fees, implementation, retries, and taxes. Savings are illustrative, not guaranteed.
Set budgets by person or team. When a premium limit is reached, switch suitable work to approved lower-cost models. Raise the limit whenever more access is needed.
Limit reached€1,250 / €1,250
Every team. Every model. The whole picture.
OpenAI24.4%Kimi18.5%DeepSeek13.4%Qwen11.8%Claude15.4%Z.ai10.0%We’re building shared rules for who can access company knowledge and what they can share with AI. Sensitive data and company IP are flagged for manager review before unapproved sharing.
customer-export.csvNeeds manager approval before sharing.
Private prompts. No training.
Where providers support it.
Less data. Clear agreements.
Traceable access.
Monitoring, incident response,
and supplier reviews.
EU-hosted model options.
Clear data flows.
Invite your teammates by email. They get access to 250+ models from 50+ providers, with the access and spending limits you choose.
Start inviting your teamKimEngineeringInvited
For ambitious teams.