GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
nkthebass
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
nkthebass
Model Tree
nkthebass/tinybrainbot-360m-a75m-v4.1-moe-base
Input Modalities
Output Modalities
Supported Functionality
| v5-Instruct (this) | v4.1-Instruct (previous) | SmolLM2-360M-Instruct | |
|---|---|---|---|
| IFEval prompt-level strict | 25.1 | 18.5 | 32.9 |
| IFEval instruction-level strict | 37.6 | 29.7 | 44.2 |
| Open SLM Leaderboard Intelligence Index | 27.82 | 23.67 | |
| (raw acc) | 26.07 |
IFEval was run with lm-eval and the chat template, greedy, with the <think> block removed before scoring. The Index uses the full HellaSwag / ARC / PIQA sets plus ArithMark-3, 0-shot.
| Benchmark (0-shot, full set) | acc_norm | (raw acc) |
|---|---|---|
| HellaSwag | 38.40 | 33.27 |
| ARC-Easy | 54.46 | 59.68 |
| ARC-Challenge | 31.74 | 27.99 |
| PIQA | 65.56 | 65.51 |
| ArithMark-3 | 57.8 | 57.3 |
Math in chat (greedy, exact final answer):
| 2-digit add | 2-digit sub | 2-digit mul | 3-digit add | 3-digit sub | 4-digit add | word problems |
|---|---|---|---|---|---|---|
| 100% | 100% | 96% | 100% | 100% | 100% | 70% (10 short problems) |
Safety (held-out prompts): it declines 93% of harmful requests and wrongly refuses 8.3% of benign ones.
python
from transformers import AutoModelForCausalLM, AutoTokenizerrepo = "nkthebass/tinybrainbot-360m-a75m-v5-instruct"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo)msgs = [{"role": "user", "content": "Leo has 30 pens. Mia has 9 fewer pens than Leo. How many pens does Mia have?"}]ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt")out = model.generate(ids, max_new_tokens=200) # defaults from generation_config.jsonprint(tok.decode(out[0][ids.shape[1]:], skip_special_tokens=True))GGUF (llama.cpp, LM Studio, Jan, Ollama): tinybrainbot-360m-a75m-v5-instruct-Q8_0.gguf (409 MB) or -F16.gguf (768 MB). The chat template is embedded, and both tokenize exactly like the transformers version.
text
llama-cli -m tinybrainbot-360m-a75m-v5-instruct-Q8_0.gguf -cnv --temp 0.4 --top-p 0.9 --repeat-penalty 1.15 --logit-bias 241-inf --logit-bias 255-infRecommended sampling: temperature 0.4, top-p 0.9, repetition penalty 1.15. The two --logit-bias flags block the byte tokens that start every emoji: the tokenizer has no emoji pieces, so the model would otherwise spell them out byte by byte.
<think> blocks are plain text, not special tokens. Strip them yourself if you only want the answer:
python
import re; answer = re.sub(r"<think>.*?</think>", "", reply, flags=re.S).strip()<think> scratchpad ("'9 fewer than' means subtraction...").Unedited sample conversations, including mistakes, are in conversation_samples.md.
| Part | Share | What |
|---|---|---|
| Chat | 40% | smol-smoltalk, multi-turn conversations, recall and follow-up dialogues, WildChat (refusals removed) |
| Math | 25% | column-routine arithmetic (wide numbers, negatives, 3x3 multiplication), cue-first word problems, generated scratchpad word problems |
| Reasoning | 25% | <think> reasoning traces, open-ended reasoning, safety-aware reasoning, with plain-chat ballast so it doesn't think on "Hi" |
| Instruction following | 10% | Tulu 3 persona instruction-following |
No benchmark training or test data was used. The GSM8K-derived sets from our earlier math model were left out on purpose.
from transformers import AutoModelForCausalLM, AutoTokenizerrepo = "nkthebass/tinybrainbot-360m-a75m-v5-instruct"tok = AutoTokenizer.from_pretrained(repo)model = AutoModelForCausalLM.from_pretrained(repo)msgs = [{"role": "user", "content": "Leo has 30 pens. Mia has 9 fewer pens than Leo. How many pens does Mia have?"}]ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt")out = model.generate(ids, max_new_tokens=200) # defaults from generation_config.jsonprint(tok.decode(out[0][ids.shape[1]:], skip_special_tokens=True))import re; answer = re.sub(r"<think>.*?</think>", "", reply, flags=re.S).strip()