Model reference · open weights

BananaMind-2-Pro

Available as managed deployment Licence fee LLMs BananaMind Text gen 1 variants 1k dl/mo

BananaMind-2-Pro is an open-weight language model from BananaMind. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.

What it is

Released byBananaMind
TypeLanguage models
TaskText gen
Parameters (lead)160M
Context3k tokens
Runs withtransformers
Released2026-08-14
Popularity1k downloads / month
LicenceCommercial licence needed

About

What BananaMind-2-Pro is

BananaMind-2-Pro is a 138,971,520-parameter decoder-only base language model trained from scratch by BananaMind. The completed pretraining run processed 99,999,449,088 tokens—the scheduled realization of a 100B-token curriculum—over 184,954 optimizer steps.

The model has a 3,072-token context window and a custom 32,768-token digit-aware byte-level BPE tokenizer. Its architecture uses grouped-query attention, QK normalization, RoPE, SwiGLU, RMSNorm, tied input/output embeddings, and a KV cache for generation.

This is a base model, not an instruction-tuned or chat model. Use continuation-style prompts and load the repository with trust_remote_code=True.

Read the full model card

Release Status

FieldValue
Release typeFinal base checkpoint
Checkpoint step184,953
Optimizer steps completed184,954
Tokens seen99,999,449,088
Training target100B tokens
Final phaseQuality finish
Training statusComplete
Evaluation statusComplete

Model Details

FieldValue
Parameters138,971,520
ArchitectureBananaMind2Pro decoder-only Transformer
Layers24
Hidden size640
Intermediate size1,920
Attention heads8
KV heads4
Head dimension80
Attention styleGrouped-query attention with QK norm
MLPSwiGLU
Position embeddingsRoPE
RoPE theta100,000
NormalizationRMSNorm
RMSNorm epsilon1e-6
Vocabulary size32,768
Context length3,072
EmbeddingsTied input/output embeddings
Generation cacheKV cache supported
Weight formatsafetensors
HF architectureBananaMind2ProForCausalLM
HF model typebananamind2_pro

Architecture Overview

The model maps tokens through a tied input embedding, 24 pre-normalized Transformer blocks, a final RMSNorm, and the tied language-model head. Each block contains causal grouped-query self-attention followed by a SwiGLU feed-forward network, with residual connections around both sublayers.

Eight query heads share four key/value heads, reducing KV-cache size while retaining multi-head query capacity. QK normalization stabilizes attention logits, and RoPE supplies position information without learned absolute-position embeddings. During autoregressive generation, each layer can reuse cached keys and values instead of recomputing the full prefix.

Evaluation

These results describe the exact final checkpoint at step 184,953. ARC Easy, ARC Challenge, PIQA, and HellaSwag use zero-shot acc_norm,none. ArithMark 3 uses length-normalized continuation accuracy, while ArithMark 2 uses raw continuation accuracy. Code Elo is the Base Bench 1.1 code-completion category; Base Bench 1.1 Elo is the complete 350-item suite.

BenchmarkBananaMind-2-ProBananaMind-2-Pro-PreviewGPT-X2.5-135MBananaMind-2-MediumGPT-2
Trainable parameters139M139M135M49.6M124M
ARC Easy53.58%51.01%51.81%43.81%39.35%
ARC Challenge27.82%27.13%29.18%25.34%22.35%
PIQA67.52%66.76%69.42%61.86%62.08%
HellaSwag42.78%39.83%40.57%32.43%31.26%
ArithMark 338.20%38.90%38.10%36.20%35.70%
ArithMark 232.08%28.60%N/A28.20%26.48%
INT Index24.9623.0425.1715.37N/A
Code Elo1407129512531034996
Base Bench 1.1 Elo1124110611061034996

The INT Index chance-normalizes HellaSwag, the mean of ARC Easy and ARC Challenge, PIQA, and ArithMark 3. ArithMark 2 is not included:

N(s, c) = 100 * (s - c) / (100 - c)
INT = [N(H,25) + N((ARC_E+ARC_C)/2,25) + N(P,50) + 0.65*N(A3,25)] / 3.65

INT Index vs Training Compute

Training compute is estimated consistently as 6 × parameters × training tokens. Using the exact 138,971,520-parameter count gives 83,382.91 PFLOPs for Pro Final at 100B tokens and 43,279.49 PFLOPs for Pro Preview at 51.904512B tokens. GPT-X2.5 uses 135M parameters, 75B tokens, and a 2,048-token context window; context length is not a separate term in this 6ND approximation.

ModelEstimated training computeINT Index
BananaMind-2-Pro Final83,382.91 PFLOPs24.96
BananaMind-2-Pro Preview43,279.49 PFLOPs23.04
GPT-X2.5-135M60,750.00 PFLOPs25.17
GPT-X2-125M56,286.75 PFLOPs23.36
GPT-X-125M11,210.56 PFLOPs19.94
Supra2-100M18,000.00 PFLOPs19.41
SmolLM-135M484,254.03 PFLOPs25.74
BananaMind-2-Medium14,867.33 PFLOPs15.37
OPT-125M135,000.00 PFLOPs13.80

Base Bench Checkpoint Progression

This progression contains 41 complete 350-item evaluations from 2.70B tokens through the final 99.999B-token checkpoint. Every point uses CUDA, bfloat16, and batch size 1 for consistency. The final batch-1 endpoint is 1132 Elo, 236/350 correct (67.43%), and 64.69% weighted accuracy. The primary comparison table above retains the separately measured batch-32 result.

Base Bench 1.1 Category Results

The final model passed 233 of 350 cases for 66.57% raw accuracy and 63.73% weighted accuracy. Its overall fixed-item rating is 1124 Elo.

CategoryEloCorrectAccuracyWeighted accuracy
Language completion157050/50100.00%100.00%
Commonsense118540/5080.00%78.86%
World knowledge116840/5080.00%77.12%
Context tracking93822/5044.00%41.72%
Quantitative88615/5030.00%28.71%
Logical reasoning105924/5048.00%44.51%
Code completion140742/5084.00%85.67%
Overall1124233/35066.57%63.73%

Base Bench was evaluated locally with CUDA, bfloat16, and batch size 32. ArithMark 3 used CUDA bfloat16 with batch size 64; ArithMark 2 used CUDA bfloat16 with batch size 1. Resul

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys bananamind-2-pro for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (bananamind-2-pro below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/chat/completions \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"bananamind-2-pro","messages":[{"role":"user","content":"Hello"}]}'

Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms