Model
Model explorer

Teuken-7B (instruct)

OPEN
OpenGPT-X · Teuken family · released Nov 26, 2024

EU-funded model pre-trained from scratch and instruction-tuned across all 24 official EU languages; Apache-2.0 commercial variant.

ReasoningCodingVisionFunction callingTool useAgentic
-3.2
Elo · rank #404
Parameters
7B
Active params
7B (dense)
Context
4K tokens
Architecture
Custom decoder-only transformer with multilingual tokenizer, GQA, RMSNorm, SwiGLU, RoPE
License
Apache-2.0
Languages
24+
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
ARC-ChallengeReasoning56.9%#101
best: Llama 3.1 405B · 96.9%
HellaSwagReasoning62.0%#140
best: Claude 3 Opus · 95.4%
MMLU (EU-21 languages)Knowledge43.0%#12
best: Llama 3.1 70B · 77.1%
TruthfulQAKnowledge50.3%#55
best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%
Run it locally
VRAM @ Q4
5 GB
VRAM @ FP16
14 GB
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
No RTX 4090-specific measurement found; a 7B model at Q4 fits comfortably and typically runs well above real-time on a 4090.
Quantizations
GGUF Q4 · AWQ · MLX 4-bit
Teuken family
Elo progression across releases
Teuken-7B (instruct)Nov 2024-3.2
Resources
HFopenGPT-X/Teuken-7B-instruct-commercial-v0.4DOCSopengpt-x.de
API price weights · each benchmark row carries its own source badge (see methodology)