Qwen3-Embedding
Qwen3-Embedding is a previous June 2025 Qwen embedding snapshot, not an unaudited 2026 MTEB crown. Rechecked 2026-08-18: leftover copy still called the 8B variant #1 on the MTEB multilingual leaderboard at 70.58. Hugging Face still lists the family, including Qwen3-Embedding-8B and Qwen3-Embedding-0.6B. Live Qwen chat on this site is Qwen3.8. Drop leftover latest / MTEB #1 / 70.58 as a 2026 census.
Compare leftover Qwen3 VL Embedding if you wanted the leftover multimodal twin, leftover text-embedding-3-small if you wanted the still-listed OpenAI cheap v3 row, and Qwen3.8 if you wanted the live Qwen chat path.
Key Features
- Previous 2025 embedding family: Leftover copy treated 0.6B / 4B / 8B as the current open embedding default.
- Not a 2026 MTEB census: Do not keep 70.58 / #1 unless you re-run a first-party card.
- Still listed on Hugging Face: Recheck the live model cards before you hard-code sizes.
- Not live Qwen chat: New Qwen chat work starts at Qwen3.8.
Use Cases
- People with a Qwen3-Embedding bookmark who need the leftover #1 copy dropped.
- People comparing leftover Qwen3-Embedding vs leftover OpenAI 3-small.
- People who were about to paste MTEB 70.58 into a 2026 deck.
Limitation: We did not re-run MTEB. Recheck first-party cards before you quote a 2026 score.
Pricing
First-party pages on 2026-08-18.
| Piece | Price | Notes |
|---|---|---|
| Qwen3-Embedding | Previous snapshot | Recheck the Hugging Face family. No invented hosted dollar row. |
| Live Qwen chat | Qwen3.8 | Qwen3.8. |
If a leftover blog still calls Qwen3-Embedding the current MTEB #1, treat this page as the correction.
Getting Started
- Open the Hugging Face Qwen embedding cards before you hard-code 0.6B / 4B / 8B.
- Recheck leftover text-embedding-3-small if you wanted a still-listed OpenAI cheap row.
- Recheck Qwen3.8 if you wanted live Qwen chat.
- Do not paste leftover MTEB #1 into a 2026 deck.
Frequently Asked Questions
Is Qwen3-Embedding still MTEB #1 in 2026?
Do not keep that leftover score unless you re-run a first-party card.
Same as Qwen3.8?
No. Qwen3.8 is the live Qwen chat path.
Did we invent a hosted dollar row?
No. Recheck first-party hosting before you quote one.
Alternatives
- text-embedding-3-small: Still-listed cheaper OpenAI v3 row.
- Qwen3 VL Embedding: Leftover multimodal embedding twin.
- Qwen3.8: Live Qwen chat if you wandered here for a current model.
Tips
- Call Qwen3-Embedding a previous 2025 snapshot, not a 2026 MTEB census.
- Recheck Hugging Face before you hard-code sizes.
- Point new Qwen chat work at Qwen3.8.
Conclusion
Qwen3-Embedding is a leftover June 2025 embedding snapshot, not a 2026 MTEB census. Start at the Hugging Face cards, then decide whether leftover text-embedding-3-small or live Qwen3.8 already covers the job.
Comments
No comments yet. Be the first to comment!
Related Tools
BGE-M3
huggingface.co/BAAI/bge-m3
2024 BAAI multilingual embedding snapshot. Rechecked: MIT card still listed. Not a 2026 MIRACL or MTEB crown.
Qwen3-VL-Embedding
huggingface.co/Qwen/Qwen3-VL-Embedding-2B
Previous Qwen3-VL embedding snapshot. Rechecked: still listed on Hugging Face. Drop leftover latest multimodal embedding crown.
Cohere Embed v3
docs.cohere.com/docs/cohere-embed
Previous Cohere embed snapshot. Rechecked: live catalog is Embed 4. Not a $0.10/1M 2026 default.
Related Insights
Grok Bot and Hermes Bot: one person finally gets a think tank and a secretariat
Grok Bot now ships with Cursor Pro+. Hermes Bot runs on a VPS. They are not smarter chat boxes. The think tank advises, the secretariat executes, and you still make the call.
Hook OpenCode Go into Codex on Windows. Do Not Open a Second Toolkit.
A ChatGPT-signed Codex desktop app still shows mostly GPT in the picker. On Windows, enable only OpenCode Go and the Grok, GLM, Kimi, DeepSeek, and MiniMax models you already pay for appear in the same selector. Keys stay local. Native GPT stays put.
What Locks Codex Is the Picker, Not the Models
You already pay for OpenCode Go, Grok, and Z.ai, but the Codex picker still shows mostly GPT. The community project codex-router does not teach another install ritual. It puts subscriptions you already bought back into the selector, keeps keys on the machine, and leaves native GPT alone.