Total runs:
1.5K
Run Growth:
-784
Growth Rate:
-65.01%
Sponsored by
Agentkit
DJLougen / Harmonic-9B
Model's Last Updated: April 10 2026
text-generation
Introduction of Harmonic-9B
Model Details of Harmonic-9B
☕ Support This Work
I'm a PhD student in visual neuroscience at the University of Toronto who also happens to spend way too much time fine-tuning, merging, and quantizing open-weight models on rented H100s and a local DGX Spark. It's a hobby that got out of hand. If my uploads have been useful to you, consider buying a PhD student a coffee. It goes a long way toward keeping these experiments running.
Harmonic-9B
Harmonic-9B is a reasoning backbone — not a general chat model, not an agent, and not a casual assistant.
It was built to think with real structure: consistent self-correction, verification steps, and multi-path exploration in every training example.
Stage 1 (this model): Heavy reasoning fine-tune on privately generated, structurally validated data. Every row passes strict quality gates. No junk, no shallow traces.
Stage 2 : Harmonic-Hermes-9B — the dedicated agent/tool-calling variant, fine-tuned on curated tool-calling and agent interaction data.
Training Approach
1,817 curated rows. The premise is simple: a small dataset with perfect structural quality beats a massive one full of noise. Rather than training on tens of thousands of unfiltered examples, every row was built to reinforce genuine reasoning behavior.
Every training row contains explicit self-correction ("wait, that's not right"), verification ("let me check by plugging back in"), and multi-path exploration ("alternatively, I could try..."). The data was generated from multiple frontier models and filtered through a custom structural quality pipeline that enforces reasoning depth, coherence, and flow patterns. 100% of rows pass all quality gates simultaneously.
A small set of everyday conversation data is mixed in to preserve the base model's conversational ability - calibrated by token ratio analysis to prevent the reasoning data from drowning out conversational patterns during training.
Training Data Quality
The reasoning data was curated using a custom structural process supervision pipeline. Key metrics:
| Metric | Value |
|---|---|
| Signal quality score | 78.7 mean (61.5 min, 90.0 max) |
| Thinking trace depth | 1,667 words average |
| Self-correction | 100% of rows (17.2 per row avg) |
| Verification | 100% of rows (10.3 per row avg) |
| Exploration | 100% of rows (6.3 per row avg) |
| Quality gate pass rate | 100% |
Every row was scored across multiple structural dimensions and only rows passing all thresholds simultaneously were included. No rows were manually curated - the pipeline is fully automated and reproducible.
How It Compares
We ran our structural quality analysis against every major public reasoning dataset used for Opus/Qwen distillation. The results:
| Dataset | Rows | Think Words | Self-Correction | Verification | Exploration | Signal Score | Gate Pass |
|---|---|---|---|---|---|---|---|
| Harmonic (ours) | 1,817 | 1,667 | 100% | 100% | 100% | 78.7 | 100% |
| Crownelius/Opus-3300x | 2,160 | 188 | 5.9% | 22.6% | 5.2% | 28.0 | 0.1% |
| nohurry/Opus-Filtered | 2,326 | 191 | 6.7% | 24.1% | 5.3% | 28.5 | 0.1% |
| TeichAI/Opus-250x | 250 | 323 | 17.2% | 26.8% | 6.8% | 24.6 | 0.4% |
| Jackrong/Qwen-700x | 633 | 6,653 | 97.5% | 97.6% | 69.8% | 75.6 | 22.7% |
| Bespoke-Stratos-17k | 16,710 | 1,322 | 88.2% | 72.7% | 59.7% | 71.7 | 49.0% |
| glaiveai/reasoning-20m | 22M+ | 799 | 64.1% | 41.4% | 37.3% | 46.2 | 12.8% |
| KingNish/reasoning-20k | 19,944 | 132 | 0.7% | 4.2% | 4.3% | 27.4 | 0.0% |
The popular Opus distillation datasets (Crownelius, nohurry, TeichAI) have less than 1% quality gate pass rate. Their thinking traces average under 200 words with near-zero self-correction. Models trained on this data learn to produce short, shallow chain-of-thought that looks like reasoning but lacks the structural behaviors that make reasoning reliable.
Jackrong and Stratos are closer competitors but still fall short on consistency. Jackrong has massive traces (6,653 words avg) but only 22.7% pass the quality gate - the thinking is verbose but wanders. Stratos has decent markers but 49% of rows still fail, meaning half the gradient updates during training push the model toward shallow patterns.
Harmonic's data is smaller by design. Every row passes. Every gradient update reinforces genuine reasoning behavior.
Reasoning Flow
Marker density measured across 20 equal segments of each thinking trace. The characteristic curve shows reasoning intensity building through the middle of the trace and peaking in the later segments as the model enters verification and self-correction before committing to an answer.
Training Configuration
base_model: Qwen/Qwen3.5-9B
dataset: 1,459 reasoning + 358 conversation rows
epochs: 1
learning_rate: 1e-4
lr_scheduler: cosine
warmup_ratio: 0.1
max_seq_length: 8192
lora_rank: 32
lora_alpha: 32
dropout: 0.05
micro_batch_size: 1
gradient_accumulation_steps: 4
weight_decay: 0.01
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("DJLougen/Harmonic-9B")
tokenizer = AutoTokenizer.from_pretrained("DJLougen/Harmonic-9B")
Reasoning format
The model uses
<think>
blocks for reasoning:
<think>
The user is asking about X. Let me consider two approaches...
Approach 1: ...
Approach 2: ...
I'll go with Approach 1 because...
Wait, I need to be careful here - this assumes Y, which may not hold.
Let me verify by checking a special case...
Yes, that confirms the result.
</think>
[Final answer here]
Intended Use
- Reasoning tasks requiring genuine multi-step thinking
- Mathematical problem-solving with self-correction
- Code analysis and generation with structured verification
- General conversation (conversational ability preserved through training design)
- Base model for Stage 2 agentic fine-tuning
Limitations
- 9B parameter model - not suitable for tasks requiring extensive world knowledge
- Reasoning traces can be verbose for simple questions
- Not optimized for tool calling — see Harmonic-Hermes-9B for agentic use
- Benchmark evaluation is ongoing
Architecture
- Base : Qwen 3.5 9B (9.65B parameters)
- Training : LoRA fine-tuning, merged into base weights
- Precision : BF16
- Context : 8192 tokens
License
Apache 2.0 - same as the base model. All training data is from Apache 2.0 or MIT licensed sources. Fully commercial use permitted.
Links
- GGUF quantizations: DJLougen/Harmonic-9B-GGUF
- Agentic variant: DJLougen/Harmonic-Hermes-9B
Runs of DJLougen Harmonic-9B on huggingface.co
26
Total runs
4
24-hour runs
6
3-day runs
5
7-day runs
-8
30-day runs
More Information About Harmonic-9B huggingface.co Model
Url of Harmonic-9B
Harmonic-9B huggingface.co Url
Provider of Harmonic-9B huggingface.co
Other API from DJLougen
Updated:April 04 2026
Total runs:
1.4K
Run Growth:
-3.4K
Growth Rate:
-234.70%
Total runs:
577
Run Growth:
8
Growth Rate:
1.39%
Total runs:
549
Run Growth:
-495
Growth Rate:
-90.16%
Total runs:
512
Run Growth:
-2.1K
Growth Rate:
-404.69%
Total runs:
471
Run Growth:
-341
Growth Rate:
-72.40%
Total runs:
421
Run Growth:
-2.9K
Growth Rate:
-699.29%
Total runs:
242
Run Growth:
-6.4K
Growth Rate:
-2660.33%
Total runs:
228
Run Growth:
-634
Growth Rate:
-278.07%
Total runs:
205
Run Growth:
27
Growth Rate:
13.17%
Total runs:
186
Run Growth:
-3.0K
Growth Rate:
-1627.42%
Total runs:
179
Run Growth:
-1.7K
Growth Rate:
-975.98%
Total runs:
179
Run Growth:
-232
Growth Rate:
-129.61%
Total runs:
170
Run Growth:
106
Growth Rate:
62.35%
Total runs:
154
Run Growth:
-2.1K
Growth Rate:
-1349.35%
Total runs:
144
Run Growth:
-290
Growth Rate:
-201.39%
Total runs:
141
Run Growth:
-1.7K
Growth Rate:
-1226.95%
Total runs:
129
Run Growth:
-3.2K
Growth Rate:
-2482.17%
Total runs:
125
Run Growth:
-170
Growth Rate:
-136.00%
Total runs:
122
Run Growth:
-984
Growth Rate:
-806.56%
Total runs:
119
Run Growth:
-1.3K
Growth Rate:
-1058.82%
Total runs:
111
Run Growth:
-254
Growth Rate:
-228.83%
Total runs:
111
Run Growth:
-797
Growth Rate:
-718.02%
Total runs:
106
Run Growth:
-930
Growth Rate:
-877.36%
Total runs:
96
Run Growth:
-303
Growth Rate:
-315.63%
Total runs:
93
Run Growth:
64
Growth Rate:
68.82%
Total runs:
92
Run Growth:
-1.3K
Growth Rate:
-1421.74%
Total runs:
91
Run Growth:
24
Growth Rate:
26.37%
Total runs:
89
Run Growth:
89
Growth Rate:
100.00%
Total runs:
83
Run Growth:
36
Growth Rate:
43.37%
Total runs:
75
Run Growth:
-123
Growth Rate:
-164.00%
Total runs:
72
Run Growth:
-192
Growth Rate:
-266.67%
Total runs:
63
Run Growth:
-169
Growth Rate:
-268.25%
Total runs:
59
Run Growth:
-960
Growth Rate:
-1627.12%
Total runs:
56
Run Growth:
-368
Growth Rate:
-657.14%
Total runs:
55
Run Growth:
-45
Growth Rate:
-81.82%
Total runs:
55
Run Growth:
-14
Growth Rate:
-25.45%
Total runs:
54
Run Growth:
54
Growth Rate:
100.00%
Total runs:
53
Run Growth:
-811
Growth Rate:
-1530.19%
Total runs:
49
Run Growth:
49
Growth Rate:
100.00%
Total runs:
48
Run Growth:
38
Growth Rate:
79.17%
Total runs:
47
Run Growth:
-492
Growth Rate:
-1046.81%
Total runs:
43
Run Growth:
-445
Growth Rate:
-1034.88%
Total runs:
37
Run Growth:
-389
Growth Rate:
-1051.35%
Total runs:
32
Run Growth:
32
Growth Rate:
100.00%
Total runs:
30
Run Growth:
-201
Growth Rate:
-670.00%
Total runs:
28
Run Growth:
-28
Growth Rate:
-100.00%
Total runs:
28
Run Growth:
-45
Growth Rate:
-160.71%
Total runs:
25
Run Growth:
-7
Growth Rate:
-28.00%
Total runs:
23
Run Growth:
-81
Growth Rate:
-352.17%
Total runs:
22
Run Growth:
-6
Growth Rate:
-27.27%
Total runs:
21
Run Growth:
-201
Growth Rate:
-957.14%
Total runs:
21
Run Growth:
-3
Growth Rate:
-14.29%
Total runs:
20
Run Growth:
-21
Growth Rate:
-105.00%
Total runs:
18
Run Growth:
-6
Growth Rate:
-33.33%
Total runs:
18
Run Growth:
-517
Growth Rate:
-2872.22%
Total runs:
18
Run Growth:
-135
Growth Rate:
-750.00%
Total runs:
12
Run Growth:
4
Growth Rate:
33.33%
Total runs:
12
Run Growth:
-10
Growth Rate:
-83.33%
Total runs:
11
Run Growth:
0
Growth Rate:
0.00%
Total runs:
8
Run Growth:
-28
Growth Rate:
-350.00%
Total runs:
5
Run Growth:
-138
Growth Rate:
-2760.00%
Total runs:
3
Run Growth:
-1
Growth Rate:
-33.33%
Total runs:
2
Run Growth:
-84
Growth Rate:
-4200.00%
Total runs:
1
Run Growth:
-105
Growth Rate:
-10500.00%
Total runs:
1
Run Growth:
-3
Growth Rate:
-300.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%