bge-large-en-v1.5 is a large-scale English sentence-embedding model released by the Beijing Academy of Artificial Intelligence (BAAI) that converts text into 1024-dimensional vectors for tasks like similarity search and clustering.
Training required a cluster of 8 NVIDIA A100 40 GB GPUs for approximately 72 hours.
ONNX Runtime or TensorRT can reduce latency to less than 5 ms per query on an RTX 3080.
q4km.ai1.2k words
6 best of 6 tracked
The 1 pick above, plus these 5. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 2 API providers; 18 quantized builds for local use; a big base for builders (105 fine-tunes, adapters and merges).
Third-party measurements only; author-reported numbers stay in Reference, marked as such.
#5 of 33 on MTEB Reranking so far
48.26. Each tick is a model on this site with a measured MTEB Reranking row; author-reported numbers stay in Reference. MAP x100 (mean of 2 tasks). 2 tasks, split=test
Everything else, folded: the full tables, every source.
Benchmarks
3 rows
Benchmarks 3 rows
measured third-party leaderboard row (3)author-reported number from the model card or eval filesNo confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Hosted endpoints observed for bge-large-en-v1.5, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.01/M.
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
bge-large-en-v1.5 is a large-scale English sentence-embedding model released by the Beijing Academy of Artificial Intelligence (BAAI) that converts text into 1024-dimensional vectors for tasks like similarity search and clustering. Benchmarked on the MTEB suite, it achieves an Amazon Polarity Classification accuracy of 92.42 % and supports export formats including PyTorch, ONNX, and SafeTensors for deployment on various hardware. The repository lists the license as unknown, requiring due diligence before commercial deployment.written, reviewed, cited
Training required a cluster of 8 NVIDIA A100 40 GB GPUs for approximately 72 hours.
ONNX Runtime or TensorRT can reduce latency to less than 5 ms per query on an RTX 3080.
BGE Large EN v1.5 is BAAI's high-performance English embedding model available on Cloudflare Workers AI for $0.204 per 1M input tokens. The model produces 1,024-dimensional dense vectors and achieved a score of 65.89 on the MTEB(eng, v2) benchmark across 41 tasks. Version 1.5 addresses similarity distribution issues from earlier releases to improve retrieval quality even when query instructions are omitted.written, reviewed, cited
BGE Large EN v1.5 is a 0.335B parameter dense model from BAAI featuring Multi-Head Attention with 24 layers and a 512 token context window. The model requires 0.7 GB of VRAM at BF16 precision and achieves 3.5K tok/s throughput on a B200 SXM GPU. Self-hosted inference on this hardware costs $4261/month, while the cheapest API deployment via together is $0.01/M output tokens.written, reviewed, cited
The model requires 98304 bytes of additional VRAM for KV-cache per token during inference.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
