Jsandero commited on
Commit
430da21
·
verified ·
1 Parent(s): 2188e81

Add M2.2 to highlights, link models

Browse files
Files changed (1) hide show
  1. README.md +11 -5
README.md CHANGED
@@ -12,17 +12,23 @@ ForgeWorks is an organisation that builds small language models. Currently best
12
  ## Highlights
13
 
14
 
15
- **ForgePlex-M2.1-10M**
16
 
17
- - Peaked at 1st 🥇 (currently 2nd) on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
 
 
 
 
 
 
18
  - The first sub-10M-parameter model to pass an Intelligence Index of 10 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) with 10.70
19
- - A +1.56 Index gain over ForgePlex-M2 (9.14), with the same core architecture.
20
 
21
- **ForgePlex-M2-9M**
22
 
23
  - Peaked at 3rd 🥉 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
24
 
25
- **ForgePlex-M1-6M**
26
 
27
  - First model trained on [Axiomic Labs](https://huggingface.co/AxiomicLabs) training framework: Trainwork.
28
  - Peaked at 23rd on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
 
12
  ## Highlights
13
 
14
 
15
+ **[ForgePlex-M2.2-10M](https://huggingface.co/ForgeWorks/ForgePlex-M2.2-10M)**
16
 
17
+ - Currently 1st 🥇 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
18
+ - The definitive M2: an Intelligence Index of 12.12 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard), post-trained for 2B tokens on top of [ForgePlex-M2.1](https://huggingface.co/ForgeWorks/ForgePlex-M2.1-10M).
19
+ - A +1.42 Index gain over [ForgePlex-M2.1](https://huggingface.co/ForgeWorks/ForgePlex-M2.1-10M) (10.70), with the same core architecture.
20
+
21
+ **[ForgePlex-M2.1-10M](https://huggingface.co/ForgeWorks/ForgePlex-M2.1-10M)**
22
+
23
+ - Peaked at 1st 🥇 (currently 3rd) on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
24
  - The first sub-10M-parameter model to pass an Intelligence Index of 10 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) with 10.70
25
+ - A +1.56 Index gain over [ForgePlex-M2](https://huggingface.co/ForgeWorks/ForgePlex-M2-9M) (9.14), with the same core architecture.
26
 
27
+ **[ForgePlex-M2-9M](https://huggingface.co/ForgeWorks/ForgePlex-M2-9M)**
28
 
29
  - Peaked at 3rd 🥉 on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.
30
 
31
+ **[ForgePlex-M1-6M](https://huggingface.co/ForgeWorks/ForgePlex-M1-6M)**
32
 
33
  - First model trained on [Axiomic Labs](https://huggingface.co/AxiomicLabs) training framework: Trainwork.
34
  - Peaked at 23rd on the [Open SLM Leaderboard](https://huggingface.co/spaces/AxiomicLabs/Open_SLM_Leaderboard) <10m parameters category.