GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
bottlecapai
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
bottlecapai
Model Tree
google/gemma-4-26B-A4B-it
Input Modalities
Output Modalities
Supported Functionality
Base google/gemma-4-26B-A4B-it against bottlecapai/ThinkingCap-gemma-4-26B-A4B-it, shown as Ours.
(Ours − Base) / Base, where each mean is taken over every question and seed of the benchmark. A saving is shown as a green ↓ percentage.We also track two ways a thinking trace can fail:
Both rates are equal-weight means across benchmarks, and they are highly correlated.
Hardware: one NVIDIA GB300 per server, without tensor parallelism. vLLM 0.26.0 with DFlash speculative decoding: draft model z-lab/gemma-4-26B-A4B-it-DFlash, 15 draft tokens.
Thinking is on (enable_thinking=True), with the base model's recommended sampling —
temperature=1.0, top_p=0.95, top_k=64 — used unchanged for Ours.
The generation cap is 65,536 tokens on every benchmark, per turn on τ²-bench. The context window is 73,728 tokens on the single-turn benchmarks, 131,072 on τ²-bench and 262,144 on AA-LCR and LongBench v2. The cap was carefully selected to bite in the cases where the base model is looping, to speed up the evals while not affecting the measured accuracy.
Twelve benchmarks: AIME 2026 (30 problems), HMMT Feb 2026 (33), HMMT Nov 2025 (30), GPQA-Diamond (198), LiveCodeBench v6 (175), MMLU-Pro (12,032), MMMLU (a fixed subset of 714 questions in each of its 14 languages, 9,996 in all), IFBench (300), RealWorldQA (765), AA-LCR (100), LongBench v2 (the 370 of its 503 questions whose prompt fits in 196,000 tokens; longer prompts are left out, not truncated) and τ²-bench (278 tasks across airline, retail and telecom).
Independent runs per benchmark: 64 seeds on AIME 2026, HMMT (Feb 2026) and HMMT (Nov 2025); 32 seeds on GPQA-Diamond and LiveCodeBench v6; 16 seeds on RealWorldQA, IFBench and τ²-bench; 8 seeds on AA-LCR and LongBench v2; 4 seeds on MMLU-Pro and MMMLU. Accuracy cells show the 95% t-interval across seeds — how much the answer moves when only the sampling seed changes, at Google's recommended temperature of 1.0.
The bf16 weights take about 52 GB. Apart from the weights, every file matches the base model, so the base model's usage examples apply with this repo's id; they include image input.
python
from transformers import AutoModelForMultimodalLM, AutoProcessorprocessor = AutoProcessor.from_pretrained("bottlecapai/ThinkingCap-gemma-4-26B-A4B-it")model = AutoModelForMultimodalLM.from_pretrained("bottlecapai/ThinkingCap-gemma-4-26B-A4B-it", dtype="auto", device_map="auto") messages = [{"role": "user", "content": "How many prime numbers are there below 100?"}]inputs = processor.apply_chat_template( messages, tokenize=True, return_dict=True, return_tensors="pt", add_generation_prompt=True, enable_thinking=True,).to(model.device)input_len = inputs["input_ids"].shape[-1]outputs = model.generate(**inputs, max_new_tokens=65536, do_sample=True, temperature=1.0, top_p=0.95, top_k=64)response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)processor.parse_response(response, prefix=inputs["input_ids"]) # splits thinking from answerThe reasoning parser returns the thinking in a separate reasoning field, and the tool-call
parser turns the model's tool calls into structured tool_calls. Both commands below run on a
single 141 GB H200 at the full 262,144-token window.
bash
vllm serve bottlecapai/ThinkingCap-gemma-4-26B-A4B-it --max-model-len 262144 \ --reasoning-parser gemma4 --enable-auto-tool-choice --tool-call-parser gemma4Our evaluation servers ran DFlash speculative decoding, which also needs the triton attention backend on the model itself:
bash
vllm serve bottlecapai/ThinkingCap-gemma-4-26B-A4B-it --max-model-len 262144 --attention-backend triton_attn \ --reasoning-parser gemma4 --enable-auto-tool-choice --tool-call-parser gemma4 \ --speculative-config '{"method":"dflash","model":"z-lab/gemma-4-26B-A4B-it-DFlash","num_speculative_tokens":15,"attention_backend":"triton_attn"}'Over every decoding step of our evaluation runs (vLLM 0.26.0, one GB300 per server), DFlash accepted on average 4.5 tokens per step of the 15 drafted with this model: 3.0 on RealWorldQA and τ²-bench, up to 6.0 on HMMT Feb 2026.
The server speaks the OpenAI Chat Completions API. Thinking is turned on per request:
python
from openai import OpenAIclient = OpenAI(base_url="http://localhost:8000/v1", api_key="-")r = client.chat.completions.create( model="bottlecapai/ThinkingCap-gemma-4-26B-A4B-it", messages=[{"role": "user", "content": [ {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}, {"type": "text", "text": "What is happening in this picture?"}, ]}], temperature=1.0, top_p=0.95, extra_body={"top_k": 64, "chat_template_kwargs": {"enable_thinking": True}},)print(r.choices[0].message.reasoning) # the thinkingprint(r.choices[0].message.content) # the answerA text-only request is the same call with a plain string as content.
Same checkpoint, chat template and license, quantized for smaller footprints:
Need even more efficiency? Our enterprise versions go further — fewer thinking tokens still, tuned to your workload, at matched accuracy on your own tasks. Built for AI labs, inference providers and enterprises running models at scale. Deployed on your infrastructure, or in the cloud and region you choose. Talk to our team
Released under the base model's own license, Apache 2.0 (see the Gemma 4 license page).
If you use this model, please cite:
bibtex
@misc{ThinkingCap-gemma-4-26B-A4B-it, title = {bottlecapai/ThinkingCap-gemma-4-26B-A4B-it}, author = {Osusky, Adam and Lindauer, Jan and Jirkovsky, Adam and Mihal, Filip and Platek, Ondrej and Herel, David and Ihnatchenko, Luka and Bartek, Vojtech and Jirak, Jiri and Kubista, Daniel and Krus, Frantisek and Mikolov, Tomas}, year = {2026},}from transformers import AutoModelForMultimodalLM, AutoProcessorprocessor = AutoProcessor.from_pretrained("bottlecapai/ThinkingCap-gemma-4-26B-A4B-it")model = AutoModelForMultimodalLM.from_pretrained("bottlecapai/ThinkingCap-gemma-4-26B-A4B-it", dtype="auto", device_map="auto") messages = [{"role": "user", "content": "How many prime numbers are there below 100?"}]inputs = processor.apply_chat_template( messages, tokenize=True, return_dict=True, return_tensors="pt", add_generation_prompt=True, enable_thinking=True,).to(model.device)input_len = inputs["input_ids"].shape[-1]outputs = model.generate(**inputs, max_new_tokens=65536, do_sample=True, temperature=1.0, top_p=0.95, top_k=64)response = processor.decode(outputs[0][input_len:], skip_special_tokens=False)processor.parse_response(response, prefix=inputs["input_ids"]) # splits thinking from answervllm serve bottlecapai/ThinkingCap-gemma-4-26B-A4B-it --max-model-len 262144 \ --reasoning-parser gemma4 --enable-auto-tool-choice --tool-call-parser gemma4vllm serve bottlecapai/ThinkingCap-gemma-4-26B-A4B-it --max-model-len 262144 --attention-backend triton_attn \ --reasoning-parser gemma4 --enable-auto-tool-choice --tool-call-parser gemma4 \ --speculative-config '{"method":"dflash","model":"z-lab/gemma-4-26B-A4B-it-DFlash","num_speculative_tokens":15,"attention_backend":"triton_attn"}'from openai import OpenAIclient = OpenAI(base_url="http://localhost:8000/v1", api_key="-")r = client.chat.completions.create( model="bottlecapai/ThinkingCap-gemma-4-26B-A4B-it", messages=[{"role": "user", "content": [ {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}, {"type": "text", "text": "What is happening in this picture?"}, ]}], temperature=1.0, top_p=0.95, extra_body={"top_k": 64, "chat_template_kwargs": {"enable_thinking": True}},)print(r.choices[0].message.reasoning) # the thinkingprint(r.choices[0].message.content) # the answer@misc{ThinkingCap-gemma-4-26B-A4B-it, title = {bottlecapai/ThinkingCap-gemma-4-26B-A4B-it}, author = {Osusky, Adam and Lindauer, Jan and Jirkovsky, Adam and Mihal, Filip and Platek, Ondrej and Herel, David and Ihnatchenko, Luka and Bartek, Vojtech and Jirak, Jiri and Kubista, Daniel and Krus, Frantisek and Mikolov, Tomas}, year = {2026},}