"Less Gemma - more human" I had a lot of fresh and creative experience with both ConicCat tunes, but they had some instability issues, that I hopefully fixed there.
Open-weight model · Text generation
OTel-2.0-LLM-31B-IT
by Farbod Tavakkoli farbodtavakkoli/OTel-2.0-LLM-31B-IT
OTel-2.0-LLM-31B-IT is an open-weight model for text generation from Farbod Tavakkoli, released under Apache License 2.0. It has 31.3B parameters and a 262,144-token context. At 16-bit it needs about 75.1 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 1.8M downloads a month.
OTel-2.0-LLM-31B-IT is a telecom-specialized instruction model post-trained from Gemma 4 31B-IT on approximately 440 billion telecom training tokens.
Parameters31.3B
Context262,144
Weights126.8 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads1.8M
Runs On
What it takes to serve OTel-2.0-LLM-31B-IT (31.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 62.5 GB | 75.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 31.3 GB | 37.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 15.6 GB | 18.8 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 9, 2026.
OTel-2.0-LLM-31B-IT on every accelerator the SAVRN Index prices, at every precision
SAVRN's Notes on OTel-2.0-LLM-31B-IT
A 31.3B-parameter model with a 262,144-token context, post-trained from Gemma 4 31B-IT on roughly 440 billion telecom tokens: OTel-2.0-LLM-31B-IT is built for network operations, standards interpretation, configuration help and retrieval-backed answering, which is why the window matters — the input is a standards document, not a ticket. Memory runs 75.1 GB at 16-bit, 37.5 GB at 8-bit, 18.8 GB at 4-bit, and every row fits one MI300X at $1.85 an hour. The Gemma4ForConditionalGeneration architecture records the base.
Apache 2.0 permits commercial deployment, modification and redistribution with notices kept, though the Gemma base's terms deserve their own read. The repository is 39 files and 126.8 GB — about twice the 62.5 GB of 16-bit weights — so confirm what the extra files hold before pulling. The card points evaluators at agentic suites such as SWE-Bench Verified and τ²-Bench with preserved-thinking mode. If your workload is not telecom, the neighboring Qwen3.6-27B offers a more general long-context run.
Model Card
By Farbod Tavakkoli, published under apache-2.0, revision de9c87a5681a.
Versioning notice: This card describes the current OTel 2.0 OSFT checkpoint. The model weights may be updated over time. For reproducible evaluation or production deployment, pin a specific model revision, checkpoint hash, or release tag rather than relying on the floating latest version.
OTel-2.0-LLM-31B-IT is a telecom-specialized instruction model post-trained from Gemma 4 31B-IT on approximately 440 billion telecom training tokens. It is the first release in the OTel 2.0 family and is designed to support telco-grade AI workflows across network operations, standards interpretation, product development, network configuration assistance, RAG, and telecom-specific question answering.
OTel 2.0 extends the original OTel effort from a RAG-oriented telecom fine-tuning release into a larger domain-adapted training program. The model was trained from a much larger standards and telecom corpus, with new data preparation coverage for direct telecom QnA, abstention, RAG, base-model-style telecom data, and general-purpose instruction-following and tool-calling examples. The current training mixture does not include telecommunications-specific MCP, tool-calling, or instruction-following examples.
Configuration
- Architecture
- Gemma4ForConditionalGeneration
- Context length (tokens)
- 262,144
- Layers
- 60
- Hidden size
- 5,376
- Feed-forward size
- 21,504
- Attention heads
- 32
- Key/value heads
- 16
- Head dimension
- 256
- Vocabulary size
- 262,144
- Sliding window (tokens)
- 1,024
- Model type
- gemma4
Identity and Version
- Repository
- farbodtavakkoli/OTel-2.0-LLM-31B-IT
- Publisher
- Farbod Tavakkoli
- Task
- Text generation
- Modality
- Text
- Library
- transformers
- Parameters
- 31.3B parameters
- Languages
- en
- Revision
- de9c87a5681a4d6dce4ea2419eec8a2aadd5e071
- First published
- 2026-07-23
- Last updated
- 2026-10-06
Files and Weights
39 files, 126.8 GB in total. The weights are 29 files totalling 126.8 GB in safetensors.
Weights29 files · 126.8 GB
Configuration5 files · 126.6 KB
Tokenizer2 files · 32.2 MB
Documentation1 file · 40.3 KB
Other1 file · 19.3 KB
Repository1 file · 2.1 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00014.safetensors | Weights | 2.8 GB | 2d8a6ceb85e0 |
| model-00001-of-00015.safetensors | Weights | 2.8 GB | 05d6cc31f1d4 |
| model-00002-of-00014.safetensors | Weights | 5.0 GB | 0572c4159bd0 |
| model-00002-of-00015.safetensors | Weights | 2.2 GB | d436d321c12c |
| model-00003-of-00014.safetensors | Weights | 4.9 GB | e8ff2c96a762 |
| model-00003-of-00015.safetensors | Weights | 4.9 GB | d2a0ecb1317a |
| model-00004-of-00014.safetensors | Weights | 4.9 GB | 196f06858c74 |
| model-00004-of-00015.safetensors | Weights | 4.9 GB | 3f708ec3ae58 |
| model-00005-of-00014.safetensors | Weights | 4.9 GB | 83cf22e97659 |
| model-00005-of-00015.safetensors | Weights | 4.9 GB | bc3a11f4b53a |
| model-00006-of-00014.safetensors | Weights | 4.8 GB | cce5e5e46a5a |
| model-00006-of-00015.safetensors | Weights | 4.8 GB | ecb8e0e7c4b4 |
| model-00007-of-00014.safetensors | Weights | 4.9 GB | b6be3d84c71d |
| model-00007-of-00015.safetensors | Weights | 4.9 GB | e0d8dd8f5e11 |
| model-00008-of-00014.safetensors | Weights | 4.9 GB | 70af83af358a |
| model-00008-of-00015.safetensors | Weights | 4.9 GB | 120ec437e88d |
| model-00009-of-00014.safetensors | Weights | 4.9 GB | 16f5cac9f9dd |
| model-00009-of-00015.safetensors | Weights | 4.9 GB | ede9fbe726b3 |
| model-00010-of-00014.safetensors | Weights | 4.9 GB | 9e75adcd906a |
| model-00010-of-00015.safetensors | Weights | 4.9 GB | cbcb624d28f9 |
| model-00011-of-00014.safetensors | Weights | 4.9 GB | dc0842a77b6c |
| model-00011-of-00015.safetensors | Weights | 4.9 GB | 5f4d60a341f7 |
| model-00012-of-00014.safetensors | Weights | 4.8 GB | b2deff051ec3 |
| model-00012-of-00015.safetensors | Weights | 4.8 GB | 1b9a619e41ba |
| model-00013-of-00014.safetensors | Weights | 4.9 GB | 7b0ac88cb864 |
| model-00013-of-00015.safetensors | Weights | 4.9 GB | 16ba7c1662a7 |
| model-00014-of-00014.safetensors | Weights | 2.7 GB | 6f853a792175 |
| model-00014-of-00015.safetensors | Weights | 2.7 GB | 7cd821f4b598 |
| model-00015-of-00015.safetensors | Weights | 1.2 GB | 217b3dbd53ad |
| config.json | Configuration | 4.4 KB | — |
| generation_config.json | Configuration | 208 B | — |
| model.safetensors.index.json | Configuration | 120.2 KB | — |
| processor_config.json | Configuration | 1.7 KB | — |
| special_tokens_map.json | Configuration | 100 B | — |
| README.md | Documentation | 40.3 KB | — |
| chat_template.jinja | Other | 19.3 KB | — |
| .gitattributes | Repository | 2.1 KB | — |
| tokenizer.json | Tokenizer | 32.2 MB | cc8d3a0ce364 |
| tokenizer_config.json | Tokenizer | 3.7 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
- Download size
- 126.8 GB
Download from Farbod Tavakkoli
Released by Farbod Tavakkoli through its official repository on Hugging Face. Read the license.
Built From
- Described by arXiv:2504.07097
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 126.8 GB |
| 16-bit | 62.5 GB |
| 8-bit | 31.3 GB |
| 4-bit | 15.6 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Compare OTel-2.0-LLM-31B-IT
Questions About OTel-2.0-LLM-31B-IT
How much GPU memory does OTel-2.0-LLM-31B-IT need?
About 75.1 GB at 16-bit and 18.8 GB at 4-bit: the weights (31.3B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run OTel-2.0-LLM-31B-IT on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use OTel-2.0-LLM-31B-IT commercially?
Yes. OTel-2.0-LLM-31B-IT is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
What is OTel-2.0-LLM-31B-IT's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
Join our Discord community. Check out the GLM-4.7 technical blog, technical report(GLM-4.5). Use GLM-4.7-Flash API services on Z.ai API Platform. One click to GLM-4.7. GLM-4.7-Flash is a 30B-A3B MoE model. As the strongest model in the 30B class, GLM-4.7-Flash offers a new option for lightweight deployment that balances performance and efficiency. Default Settings (Most Tasks) For multi-turn agentic tasks (τ²-Bench and Terminal Bench 2), please turn on Preserved Thinking mode. Terminal Bench, SWE Bench Verified τ^2-Bench For τ^2-Bench evaluation, we added an additional prompt to the Retail and Telecom user interaction to avoid failure modes caused by users ending the interaction…
Blockwise FP8 quantization of zai-org/GLM-4.7-Flash: e4m3 weights with one scale per 128×128 block and dynamic activation scaling, about half the size of the BF16 original (30 GiB vs 58 GiB). See the original model card for model details. gsm8k (5-shot, full test set) on SGLang with MTP speculative decoding: 0.809 for FP8 vs 0.819 for BF16; average accept length 2.41 vs 2.43. Attention tensor parallelism must be 1 or 2: at 4, each rank's kvbproj shard (2240 rows) is not a multiple of the 128-row block. For more GPUs, use DP attention with expert parallelism, for example --tp-size 4 --dp-size 4 --enable-dp-attention --ep-size 4 --moe-a2a-backend deepep --cuda-graph-max-bs-decode 128. MIT…
As of 2025-10-08, create a fresh Python environment and run: For more details, refer to vLLM Official Qwen3-VL Guide Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date. This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities. Available in Dense and MoE architectures that scale from edge to cloud, with Instruct and reasoning‑enhanced Thinking editions for flexible, on‑demand deployment. Text Understanding on par with pure LLMs: Seamless text–vision…
September 2025 \- December 2025 The post-training data has a cutoff date of November 28, 2025\. The pre-training data has a cutoff date of June 25, 2025\. Nemotron-3-Nano-30B-A3B-BF16 is a large language model (LLM) trained from scratch by NVIDIA, and designed as a unified model for both reasoning and non-reasoning tasks. It responds to user queries and tasks by first generating a reasoning trace and then concluding with a final response. The model's reasoning capabilities can be configured through a flag in the chat template. If the user prefers the model to provide its final answer without intermediate reasoning traces, it can be configured to do so, albeit with a slight decrease in…