⚡

Developer 3s Key Decision Metrics

TL;DR Verdict
🏆 Benchmark Rank72s 必看
Official card: LiveCodeBench v6 85.9
Full Leaderboard→
💰 Cost & Plans比价
API & Subscription
Compare Plans↗
🛠️ Ecosystem & IDERules
kolibri · aleph-alpha · open-weight
Setup IDE Rules↗

On 3 Oct 2026, Aleph Alpha released open-weight Kolibri-1 (HF, Apache-2.0): 78B MoE / ~3.46B active, bilingual DE/EN, 262k native context (1M extrapolated), reasoning effort + tools. Official card: LiveCodeBench v6 85.9, SWE-Bench Verified 66.4, TerminalBench 2.1 27.7. Serve via aleph-alpha-inference + vLLM.

Key Takeaways

  • ✓Shipped 2026-10-03; weights Aleph-Alpha/Kolibri-1 Apache-2.0; blog on aleph-alpha.com
  • ✓78B total / ~3.46B active; ~78GB FP8; native 262k context (validated to 1M)
  • ✓Official card: LiveCodeBench v6 85.9 · SWE-Verified 66.4 · TerminalBench 2.1 27.7 · HumanEval+ 92.7
  • ✓Serve: aleph-alpha-inference + vllm serve Aleph-Alpha/Kolibri-1 with kolibri1 parsers
  • ✓Sovereign DE/EN on-prem focus — not a Terminal-Bench leader vs denser peers
🧭

Not enough VRAM? Compare cloud API and self-hosting costs

Compare 29+ dev plans & simulate token costs vs $20/mo subscriptions

⚔️ Compare Model→💰 Cloud cost calculator↗
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
📝
Aleph Alpha blog — Kolibri Has Landed
aleph-alpha.com/en/blog/kolibri-has-landed-a-sovereign-open-weight-model/
↗
🚀
Hugging Face — Aleph-Alpha/Kolibri-1
huggingface.co/Aleph-Alpha/Kolibri-1
↗
📖
Kolibri tech report (PDF)
aleph-alpha.com/downloads/tech-report.pdf
↗
🐙
GitHub — Aleph-Alpha/aleph-alpha-inference
github.com/Aleph-Alpha/aleph-alpha-inference
↗
📖
EU GPAI Code of Practice (referenced)
digital-strategy.ec.europa.eu/en/policies/contents-code-gpai
↗
🔬

In-Depth Technical Analysis

Why it matters

Aleph Alpha released Kolibri / Kolibri-1 on 3 Oct 2026 (German Unity Day): an open-weight bilingual (DE/EN) MoE under Apache-2.0 on Hugging Face. 78B total / ~3.46B active, built for regulated on-prem and agentic tool use—not a closed demo.

Architecture

50-layer MoE (384 experts, top-6 + shared), FP8 footprint ~78 GB. Native context 262k with validated extrapolation to 1M; mostly 512-token sliding-window attention. Reasoning effort none|low|medium|high plus Hermes-style tools via aleph-alpha-inference + vLLM (vllm serve Aleph-Alpha/Kolibri-1).

Benchmarks (official card only)

LiveCodeBench v6 85.9, SWE-Bench Verified 66.4, TerminalBench 2.1 27.7, HumanEval+ 92.7, AIME 2026 96.0, GPQA Diamond 84.3, Overall EN/DE 75.5/70.8. Strong DE/EN + grounding story; Terminal-Bench is not leadership vs denser peers. Full tables: tech report.

Get started

Weights on HF; pip install 'aleph-alpha-inference>=1' then the documented vLLM serve flags. Sampling: T=1.0, top_p=0.97, top_k=128. Blog: https://aleph-alpha.com/en/blog/kolibri-has-landed-a-sovereign-open-weight-model/

Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

⚔️ Launch VS Compare→💳 View Plans↗