LLMTR / llmtr/gemma-4
Gemma 4 - access through LLMTR
A first-party model hosted in Turkey, so prompts stay on LLMTR infrastructure rather than being routed to a third party. It supports a 128K context, image input and function calling, which makes it a local default for internal assistants, document analysis and tool-using agent flows. Send images as base64 `data:image/...` URLs; remote https URLs are not supported. Repeated prompts return from the prefix cache at a discount and far faster. The model runs on our own hardware and that capacity is shared. On an idle box a 256-token answer comes back in about 8 seconds (roughly 30 tokens per second); at busy times the same work can take several times longer, and the production average is around 7 tokens per second. Long prompts add a cost of their own: a 20,000-token prompt that misses the prefix cache spends about 45 seconds before generation even starts, while a repeat of the same prompt returns from the cache both cheaper and far faster. Use `stream: true`, cap the answer with `max_tokens` when you do not need the full length, and raise your client read timeout to at least 300 seconds; the 60-second default in most HTTP clients is not enough for this model.
Technical specifications
| Canonical ID | llmtr/gemma-4 |
|---|---|
| Provider | LLMTR |
| Context window | 131,072 tokens |
| Operations | CHAT_COMPLETIONS |
| Modalities | text, image |
Pricing
An 8% platform margin applies to credit top-ups; model usage prices are not separately marked up.
| Operation | Metric | Unit | Price |
|---|---|---|---|
| CHAT_COMPLETIONS | INPUT_TEXT | PER_1M_TOKENS | $2.00 |
| CHAT_COMPLETIONS | CACHE_READ | PER_1M_TOKENS | $0.500000 |
| CHAT_COMPLETIONS | OUTPUT_TEXT | PER_1M_TOKENS | $5.00 |
Example usage
With existing OpenAI SDK flows, change only the base URL and model identifier.
curl https://llmtr.com/v1/chat/completions -H "Authorization: Bearer llmtr-your_key" -H "Content-Type: application/json" -d '{"model":"llmtr/gemma-4","messages":[{"role":"user","content":"Hello"}]}'Guides about this model
- Turkey-hosted LLM models: what sits beside EVREN in the catalog - EVREN is not the only Turkey-hosted option. This is every Turkey row in the catalog in one table: which context, which price, which input type, and which rows declare tool calling.
- Adding the LLMTR gateway entry to OpenClaude and how discovery works - OpenClaude defines LLMTR as its own gateway record with a dedicated base URL, a dedicated credential variable and a hybrid catalog fed by seed entries. This guide walks that record field by field and explains how the model list fills up.
- The LLMTR provider block in OpenCode configuration and model naming - In OpenCode the first-party llmtr routes arrive preloaded, while every other catalog identifier is added to the configuration file by hand. This guide covers the connect flow, the shape of the provider block and model naming.
- OpenClaw LLMTR plugin: package install and manifest walkthrough - OpenClaw does not bake the LLMTR provider into its core. It carries the integration as a package under the extensions directory with its own manifest and model catalog, and this guide unpacks both.
- Cline with LLMTR: provider setup inside the editor panel - Cline picks a provider from a settings panel inside the editor and stores the chosen model separately for Plan and Act. This guide walks the panel flow, the request the handler builds and the SDK registration.
- PentestCode LLMTR provider setup for authorised engagements - Working from the actual files on the PentestCode LLMTR branch, this guide covers key resolution, the seeded and live model catalog, and the request transform that normalises reasoning fields per model, all inside an authorised engagement scope.
- Strix LLMTR setup: model prefix, dedupe and the reporting chain - Built from the real files on two separate Strix branches, this guide walks through model prefix resolution, why the dedupe model needs its own endpoint, and which Turkey-hosted rows actually support the tool calls an agent loop needs.
- DeepSeek Harness LLMTR plugin: the LLM seam and catalogue - In DeepSeek Harness a provider integration is not written into the core; it arrives as a plugin attached to the ctx.llm seam. The model list is not bundled either, it is read from the endpoint.
- Running OpenBot on LLMTR: one base URL, two agent runtimes - OpenBot is a self-hosted system that can give every agent a computer of its own. The integration is a matter of environment values rather than code, and a single base URL redirects three separate readers at once.
- Connecting Open WebUI to LLMTR: keeping the chat model list usable - Open WebUI treats a gateway catalog as if every entry were a chat model. This guide sets up the connection, inspects the catalog without a key, and narrows the picker to models that can actually hold a conversation.
- The models.dev LLMTR provider entry and metadata sync - models.dev is not a service but an open metadata database holding model ids and capabilities in TOML files. We compare the current LLMTR entry against the live catalog to see which rows are missing and which one is retired.
Related models
- Qwen 3.6 35B-A3B llmtr/qwen3-6-35b
- Trendyol Asure 12B llmtr/trendyol-asure-12b
- Magibu 11B v8 llmtr/magibu-11b-v8
- Muse Glimmer 30B (Turkey) llmtr/muse-glimmer-30b-tr
- EmbeddingGemma 300M llmtr/embeddinggemma-300m
- Qwen 3.5 4B llmtr/qwen3-5-4b
- GPT-6 Astra openai/gpt-6-astra
- GPT-6 Sol openai/gpt-6-sol