All tasks / Models
zembed-1
ZeroEntropy's 4B multilingual text embedding model, sold by API with non-commercial open weights.
- ZeroEntropy
- Open weights
- Released Mar 2026
Where it ranks
- BEmbeddings & retrievalModels · Local · 56.6% on MTEB(Multilingual, v2)
Facts
- Vendor
- ZeroEntropy
- Family
- zembed
- Released
- Mar 2026
- Weights
- Open: download and run it yourself
- License
- CC-BY-NC-4.0
- Parameters
- 4.02B
- GPU memory to run locally
- ~8 GB (Board lists ~7.7 GB memory use (4.02B parameters, bf16).)
- API price
- $0.05 in / $0 out per 1M tokens (input only)
- Access
- Free (open weights)
- Last verified
- 19 Oct 2026
Price from zeroentropy.dev, 19 Oct 2026.
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Agentset Embedding Model Leaderboard | 1,590 | zembed-1ELO from GPT-5-judged pairwise comparisons; rank 2 of 18; win rate 59.2%; page says 'Last updated: February 15, 2026'. | 15 Feb 2026 | agentset.ai |
| MTEB(Multilingual, v2) | 56.6% | zeroentropy/zembed-1Mean (Task) over all 131 tasks; board rank 65 (Borda); zero-shot share not declared on the board (-1); read from the leaderboard's public API. | 19 Oct 2026 | mteb-leaderboard-backend.hf.space |
Where you can use it
No tracked tool lists this model yet; as open weights it runs in local runtimes.