Total runs:
44.0K
Run Growth:
31.6K
Growth Rate:
71.78%
Sponsored by
APIMart
duyntnet / DeepCoder-14B-Preview-imatrix-GGUF
Model's Last Updated: April 17 2025
text-generation
Introduction of DeepCoder-14B-Preview-imatrix-GGUF
Model Details of DeepCoder-14B-Preview-imatrix-GGUF
Quantizations of https://huggingface.co/agentica-org/DeepCoder-14B-Preview
Open source inference clients/UIs
Closed source inference clients/UIs
- LM Studio
- Backyard AI
- More will be added...
From original readme
DeepCoder-14B-Preview is a code reasoning LLM fine-tuned from DeepSeek-R1-Distilled-Qwen-14B using distributed reinforcement learning (RL) to scale up to long context lengths. The model achieves 60.6% Pass@1 accuracy on LiveCodeBench v5 (8/1/24-2/1/25), representing a 8% improvement over the base model (53%) and achieving similar performance to OpenAI's o3-mini with just 14B parameters.
Data
Our training dataset consists of approximately 24K unique problem-tests pairs compiled from:
- Taco-Verified
- PrimeIntellect SYNTHETIC-1
- LiveCodeBench v5 (5/1/23-7/31/24)
Training Recipe
Our training recipe relies on an improved version of GRPO (GRPO+) and iterative context lengthening, introduced in DeepScaleR.
GRPO+
We enhance the original GRPO algorithm with insights from DAPO to enable more stable training:
- Offline Difficulty Filtering: DAPO employs online dynamic sampling, discarding both entirely correct and entirely incorrect samples on the fly. While this helps maintain a more stable effective batch size, it introduces significant runtime overhead due to rejection sampling. Instead, we perform offline difficulty filtering on a subset of coding problems to ensure the training dataset remains within a suitable difficulty range.
- No Entropy Loss: We observed that including an entropy loss term often led to instability, with entropy growing exponentially and ultimately collapsing training. To mitigate this, we eliminate the entropy loss entirely.
- No KL Loss: Eliminating KL loss prevents the LLM from staying within trust region of the original SFT model. This removal also obviates the need to compute log probabilities for the reference policy, thereby accelerating training.
- Overlong Filtering (from DAPO): To preserve long-context reasoning, we mask the loss for truncated sequences. This technique enables DeepCoder to generalize to 64K-context inference despite being trained with a 32K context.
- Clip High (from DAPO): By increasing the upper bound in GRPO/PPO’s surrogate loss, we encourage more exploration and more stable entropy.
Iterative Context Lengthening
Our original
Deepscaler-1.5B-Preview
scaled long context training from 8K→16K→24K, achieving 33→38→43% on AIME respectively. Similarly,
Deepcoder-14B-Preview
is trained on 16K→32K, achieving 54→58% on LiveCodeBench (v5).
DeepCoder-14B-Preview
successfully generalizes to longer contexts when evaluated at 64K context, reaching 60.6%.
DeepCoder generalizes better to long contexts than the base distilled model, due to DAPO's overlong filtering. However, it's longer responses are often truncated when the max length is capped at 16K, which can lower its scores.
| Model | 16K | 32K | 64K |
|---|---|---|---|
| DeepCoder-14B-Preview | 45.6 | 57.9 | 60.6 |
| DeepSeek-R1-Distill-Qwen-14B | 50.2 | 53.0 | 53.0 |
A more detailed description of the training recipe can be found in our blog post .
Evaluation
We evaluate
Deepcoder-14B-Preview
on various coding benchmarks, including LiveCodeBench (LCBv5), Codeforces, and HumanEval+.
| Model | LCB (v5)(8/1/24-2/1/25) | Codeforces Rating | Codeforces Percentile | HumanEval+ |
|---|---|---|---|---|
| DeepCoder-14B-Preview (ours) | 60.6 | 1936 | 95.3 | 92.6 |
| DeepSeek-R1-Distill-Qwen-14B | 53.0 | 1791 | 92.7 | 92.0 |
| O1-2024-12-17 (Low) | 59.5 | 1991 | 96.1 | 90.8 |
| O3-Mini-2025-1-31 (Low) | 60.9 | 1918 | 94.9 | 92.6 |
| O1-Preview | 42.7 | 1658 | 88.5 | 89 |
| Deepseek-R1 | 62.8 | 1948 | 95.4 | 92.6 |
| Llama-4-Behemoth | 49.4 | - | - | - |
Serving DeepCoder
Our model can be served using popular high-performance inference systems:
- vLLM
- Hugging Face Text Generation Inference (TGI)
- SGLang
- TensorRT-LLM
All these systems support the OpenAI Chat Completions API format.
Usage Recommendations
Our usage recommendations are similar to those of R1 and R1 Distill series:
- Avoid adding a system prompt; all instructions should be contained within the user prompt.
-
temperature = 0.6 -
top_p = 0.95 -
This model performs best with
max_tokensset to at least64000
Runs of duyntnet DeepCoder-14B-Preview-imatrix-GGUF on huggingface.co
825
Total runs
13
24-hour runs
-33
3-day runs
-42
7-day runs
314
30-day runs
More Information About DeepCoder-14B-Preview-imatrix-GGUF huggingface.co Model
Url of DeepCoder-14B-Preview-imatrix-GGUF
DeepCoder-14B-Preview-imatrix-GGUF huggingface.co Url
Provider of DeepCoder-14B-Preview-imatrix-GGUF huggingface.co
Other API from duyntnet
Updated:August 09 2025
Total runs:
4.0K
Run Growth:
841
Growth Rate:
21.21%
Total runs:
3.5K
Run Growth:
1.3K
Growth Rate:
36.27%
Total runs:
3.3K
Run Growth:
579
Growth Rate:
17.35%
Total runs:
3.3K
Run Growth:
1.8K
Growth Rate:
56.57%
Total runs:
2.8K
Run Growth:
639
Growth Rate:
22.69%
Total runs:
2.7K
Run Growth:
2.3K
Growth Rate:
87.87%
Total runs:
2.6K
Run Growth:
693
Growth Rate:
26.22%
Total runs:
2.6K
Run Growth:
411
Growth Rate:
15.65%
Total runs:
2.4K
Run Growth:
1.5K
Growth Rate:
64.12%
Total runs:
2.3K
Run Growth:
445
Growth Rate:
19.59%
Total runs:
2.2K
Run Growth:
353
Growth Rate:
15.81%
Total runs:
2.2K
Run Growth:
1.6K
Growth Rate:
73.44%
Total runs:
2.2K
Run Growth:
631
Growth Rate:
28.07%
Total runs:
2.2K
Run Growth:
1.3K
Growth Rate:
57.10%
Total runs:
2.2K
Run Growth:
426
Growth Rate:
18.72%
Total runs:
2.1K
Run Growth:
447
Growth Rate:
21.21%
Total runs:
2.1K
Run Growth:
455
Growth Rate:
21.93%
Total runs:
2.0K
Run Growth:
302
Growth Rate:
14.76%
Total runs:
2.0K
Run Growth:
851
Growth Rate:
43.11%
Total runs:
2.0K
Run Growth:
361
Growth Rate:
18.44%
Total runs:
1.9K
Run Growth:
1.3K
Growth Rate:
68.78%
Total runs:
1.9K
Run Growth:
419
Growth Rate:
21.97%
Total runs:
1.9K
Run Growth:
967
Growth Rate:
51.19%
Total runs:
1.8K
Run Growth:
694
Growth Rate:
37.55%
Total runs:
1.8K
Run Growth:
571
Growth Rate:
30.32%
Total runs:
1.8K
Run Growth:
501
Growth Rate:
27.54%
Total runs:
1.8K
Run Growth:
367
Growth Rate:
20.96%
Total runs:
1.7K
Run Growth:
673
Growth Rate:
39.24%
Total runs:
1.7K
Run Growth:
622
Growth Rate:
36.35%
Total runs:
1.7K
Run Growth:
392
Growth Rate:
23.10%
Total runs:
1.7K
Run Growth:
417
Growth Rate:
24.53%
Total runs:
1.6K
Run Growth:
895
Growth Rate:
54.77%
Total runs:
1.6K
Run Growth:
562
Growth Rate:
34.80%
Total runs:
1.6K
Run Growth:
811
Growth Rate:
51.69%
Total runs:
1.5K
Run Growth:
887
Growth Rate:
58.16%
Total runs:
1.5K
Run Growth:
441
Growth Rate:
27.77%
Total runs:
1.5K
Run Growth:
32
Growth Rate:
2.12%
Total runs:
1.5K
Run Growth:
857
Growth Rate:
57.48%
Total runs:
1.5K
Run Growth:
955
Growth Rate:
64.05%
Total runs:
1.5K
Run Growth:
796
Growth Rate:
54.11%
Total runs:
1.4K
Run Growth:
857
Growth Rate:
60.10%
Total runs:
1.4K
Run Growth:
1.2K
Growth Rate:
82.18%
Total runs:
1.4K
Run Growth:
1.1K
Growth Rate:
79.27%
Total runs:
1.4K
Run Growth:
309
Growth Rate:
21.14%
Total runs:
1.4K
Run Growth:
1.1K
Growth Rate:
82.91%
Total runs:
1.4K
Run Growth:
748
Growth Rate:
55.04%
Total runs:
1.3K
Run Growth:
521
Growth Rate:
38.65%
Total runs:
1.3K
Run Growth:
1.1K
Growth Rate:
81.40%
Total runs:
1.3K
Run Growth:
1.0K
Growth Rate:
81.44%
Total runs:
1.3K
Run Growth:
810
Growth Rate:
61.64%
Total runs:
1.3K
Run Growth:
658
Growth Rate:
50.65%
Total runs:
1.3K
Run Growth:
1.1K
Growth Rate:
86.40%
Total runs:
1.3K
Run Growth:
927
Growth Rate:
73.05%
Total runs:
1.3K
Run Growth:
191
Growth Rate:
15.17%
Total runs:
1.3K
Run Growth:
1.2K
Growth Rate:
98.17%
Total runs:
1.2K
Run Growth:
216
Growth Rate:
17.42%
Total runs:
1.2K
Run Growth:
321
Growth Rate:
25.99%
Total runs:
1.2K
Run Growth:
197
Growth Rate:
16.28%
Total runs:
1.2K
Run Growth:
312
Growth Rate:
25.59%
Total runs:
1.2K
Run Growth:
679
Growth Rate:
55.66%
Total runs:
1.2K
Run Growth:
405
Growth Rate:
33.69%
Total runs:
1.2K
Run Growth:
928
Growth Rate:
77.72%
Total runs:
1.2K
Run Growth:
590
Growth Rate:
49.66%
Total runs:
1.2K
Run Growth:
821
Growth Rate:
70.11%
Total runs:
1.2K
Run Growth:
763
Growth Rate:
66.23%
Total runs:
1.1K
Run Growth:
891
Growth Rate:
77.48%
Total runs:
1.1K
Run Growth:
798
Growth Rate:
69.51%
Total runs:
1.1K
Run Growth:
970
Growth Rate:
84.86%
Total runs:
1.1K
Run Growth:
1.0K
Growth Rate:
88.44%
Total runs:
1.1K
Run Growth:
234
Growth Rate:
20.62%
Total runs:
1.1K
Run Growth:
749
Growth Rate:
68.21%
Total runs:
1.1K
Run Growth:
789
Growth Rate:
70.07%
Total runs:
1.1K
Run Growth:
738
Growth Rate:
65.60%
Total runs:
1.1K
Run Growth:
390
Growth Rate:
35.58%
Total runs:
1.1K
Run Growth:
382
Growth Rate:
34.95%
Total runs:
1.1K
Run Growth:
751
Growth Rate:
67.84%
Total runs:
1.1K
Run Growth:
848
Growth Rate:
77.37%
Total runs:
1.1K
Run Growth:
612
Growth Rate:
56.35%
Total runs:
1.1K
Run Growth:
799
Growth Rate:
73.91%
Total runs:
1.1K
Run Growth:
211
Growth Rate:
20.17%
Total runs:
1.1K
Run Growth:
985
Growth Rate:
84.77%
Total runs:
1.1K
Run Growth:
464
Growth Rate:
45.31%
Total runs:
1.1K
Run Growth:
788
Growth Rate:
74.55%
Total runs:
1.1K
Run Growth:
387
Growth Rate:
36.13%
Total runs:
1.1K
Run Growth:
774
Growth Rate:
73.64%
Total runs:
1.0K
Run Growth:
316
Growth Rate:
29.62%
Total runs:
1.0K
Run Growth:
659
Growth Rate:
63.86%
Total runs:
1.0K
Run Growth:
800
Growth Rate:
77.59%
Total runs:
1.0K
Run Growth:
614
Growth Rate:
60.26%
Total runs:
1.0K
Run Growth:
598
Growth Rate:
59.38%
Total runs:
996
Run Growth:
312
Growth Rate:
31.33%
Total runs:
994
Run Growth:
379
Growth Rate:
38.13%
Total runs:
991
Run Growth:
305
Growth Rate:
30.78%
Total runs:
989
Run Growth:
660
Growth Rate:
66.73%
Total runs:
988
Run Growth:
586
Growth Rate:
59.31%
Total runs:
987
Run Growth:
695
Growth Rate:
70.49%
Total runs:
975
Run Growth:
588
Growth Rate:
60.31%
Total runs:
962
Run Growth:
416
Growth Rate:
43.24%
Total runs:
955
Run Growth:
754
Growth Rate:
78.95%