NV-Embed-v2
NV-Embed-v2 is NVIDIA's 2024 embedding snapshot, not a 2026 MTEB crown. Rechecked 2026-08-17: huggingface.co/nvidia/NV-Embed-v2 is still listed. The card we opened still talks about retrieval and a long context around 32768 tokens, not a fresh "MTEB #1 / 69.3" board. Drop leftover latest NVIDIA embedder / MTEB rank #1 / 69.3 average / 4096 as the 2026 spec.
Compare text-embedding-3-large if you wanted a still-listed OpenAI v3 row, and BGE-M3 if you wanted another 2024 open embedding card.
Key Features
- Previous NVIDIA card: The HF page still hosts NV-Embed-v2. Recheck that card for license, context, and files.
- Not a 2026 leaderboard: 69.3 / rank #1 stays off the table unless you re-run a first-party MTEB board.
- Context is not assumed 4096: The card we opened mentioned a much longer leftover context. Quote the live card, not the old catalog.
- No invented API dollar: Recheck NVIDIA API or NIM before you quote a hosted row.
Use Cases
- People with a 2024 NV-Embed-v2 bookmark who need the catalog corrected.
- People comparing leftover NVIDIA embed vs billed OpenAI v3.
- People who were about to paste MTEB #1 into a 2026 deck.
Limitation: We did not re-run MTEB or call a live NVIDIA embedding endpoint.
Pricing
First-party pages on 2026-08-17.
| Piece | Price | Notes |
|---|---|---|
| NV-Embed-v2 weights | HF card still listed | Recheck license and files on the card. |
| Hosted NVIDIA API | Recheck NVIDIA / NIM | Do not invent a per-1M dollar. |
If a leftover blog still calls v2 the current MTEB #1, treat this page as the correction.
Getting Started
- Open huggingface.co/nvidia/NV-Embed-v2 before you download weights.
- Recheck text-embedding-3-large if you wanted a billed OpenAI v3 row.
- Recheck BGE-M3 if you wanted another 2024 open card.
- Do not paste MTEB 69.3 into a deck.
Frequently Asked Questions
Is NV-Embed-v2 still MTEB #1?
Not as a 2026 census.
Is it NVIDIA's latest embedder?
Not as a live catalog claim. Recheck NVIDIA docs for newer rows.
Same as BGE-M3?
No. BGE-M3 is a 2024 BAAI multilingual card. This slug is NVIDIA's 2024 embed snapshot.
Alternatives
- text-embedding-3-large: Still-listed OpenAI v3 large row.
- BGE-M3: 2024 open multilingual card.
- text-embedding-3-small: Cheaper listed OpenAI v3 row.
Tips
- Call NV-Embed-v2 a 2024 snapshot.
- Quote context and license from the live HF card.
- Do not invent a 2026 MTEB rank.
Conclusion
NV-Embed-v2 is a 2024 NVIDIA embedding snapshot that is still on Hugging Face, not a 2026 MTEB #1 census. Start at huggingface.co/nvidia/NV-Embed-v2, then decide whether an OpenAI v3 row already covers the retrieval path you need.
Comments
No comments yet. Be the first to comment!
Related Tools
text-embedding-3-large
developers.openai.com/api/docs/models/text-embedding-3-large
2024 OpenAI embedding snapshot. Rechecked: still listed. Not an unaudited 2026 flagship or MIRACL census.
BGE-M3
huggingface.co/BAAI/bge-m3
2024 BAAI multilingual embedding snapshot. Rechecked: MIT card still listed. Not a 2026 MIRACL or MTEB crown.
text-embedding-3-small
openai.com/api/pricing
Previous OpenAI embedding snapshot. Rechecked: still listed at $0.020/1M. Not a 2026 embedding crown.
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.