ColSmol 500M measured on 4 of 203 Multimodal benchmarks4
BenchmarkByScoreRank
No benchmark matches
self-host onlyprice
77.1MIEB lite
Jan 2025released
8K tokensMax input
enLanguages
Ranking?
Full ranking →···
42NanoVDR S Multi77.543ColSmol 500M77.144Colpali v1.176.1Strong at
Alternatives?
ModelScoreCostPriceWhat changes
BetterJina Embeddings v4unstatedOpen+13%—$0.05Languages: not verified+1DOpenArgus Colqwen3.5 9B v0unstated+16%——Memory (4-bit): 0.3 GB → 5.3 GBVBetterVoyage Multimodal 3+4%—$0.12Languages: not verified
BetterNemotron Colembed VL 8B v2Open+16%——Languages: not verified+1DBetterArgus Colqwen3.5 4B v0unstatedOpen+16%——Memory (4-bit): 0.3 GB → 2.8 GBModelScoreCostPriceWhat changes
Nothing here yet.
ModelScoreCostPriceWhat changes
Jina Embeddings v4unstatedOpen+13%—$0.05Languages: not verified+1VVoyage Multimodal 3+4%—$0.12Languages: not verifiedDArgus Colqwen3.5 9B v0unstatedOpen+16%——Memory (4-bit): 0.3 GB → 5.3 GBNemotron Colembed VL 8B v2Open+16%——Languages: not verified+1DArgus Colqwen3.5 4B v0unstatedOpen+16%——Memory (4-bit): 0.3 GB → 2.8 GBModelScoreCostPriceWhat changes
DArgus Colqwen3.5 9B v0unstatedOpen+16%——Memory (4-bit): 0.3 GB → 5.3 GBNemotron Colembed VL 8B v2Open+16%——Languages: not verified+1DArgus Colqwen3.5 4B v0unstatedOpen+16%——Memory (4-bit): 0.3 GB → 2.8 GBvVultronRetrieverPrime Qwen3.5 8BOpen+15%——Memory (4-bit): 0.3 GB → 5 GBvVultronRetrieverCore Qwen3.5 4.5BOpen+15%——Memory (4-bit): 0.3 GB → 2.7 GBModelScoreCostPriceWhat changes
vColqwen2.5 v0.2Open+12%——Memory (4-bit): 0.3 GB → 1.8 GBvColqwen2 v1.0Open+11%——Memory (4-bit): 0.3 GB → 1.3 GBvColpali v1.3Open+3%——Memory (4-bit): 0.3 GB → 1.8 GBvColpali v1.2Open+2%——Memory (4-bit): 0.3 GB → 1.8 GBvColpali v1.1Open-1%——Memory (4-bit): 0.3 GB → 1.8 GBvColSmol 256MOpen-5%——All vidore models →Run it yourself
estimateFits on✓ 8 GB (phone, laptop) · full quality✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceApache-2.0free for commercial use
Benchmarks
in Embeddings & RetrievalFAQ
How much does ColSmol 500M cost?›
There is no per-unit API price: self-host only.
Is ColSmol 500M free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is ColSmol 500M, and what is it best at?›
#43 of 89 in Multimodal.
Is ColSmol 500M better than NanoVDR S Multi?›
On MIEB lite · Document Understanding · English, ColSmol 500M scores 77.1 vs 77.5 — about the same.
When was ColSmol 500M released? Is there a newer version?›
Released 2025-01-22. It is the current version in its line.
Is ColSmol 500M open source?›
Yes, the weights are open (Apache-2.0 licence).
Can I run ColSmol 500M locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
What is ColSmol 500M?›
ColSmol 500M is an open-weights embeddings & retrieval model from vidore, released Jan 2025.
What are the best alternatives to ColSmol 500M?›
Jina Embeddings v4 (better), Argus Colqwen3.5 9B v0 (open weights).