Gemma 4 26B A4B
INT8
- Context
- 131k
- Input /M
- $0.0441/M
- Output /M
- $0.231/M
- Cache read /M
- N/A
- Cache hit
- 0%
- TTFT
- 1.8s
- Throughput
- 21 tps
Darkbloom
GLB
15 models available
Darkbloom says ordinary coordinator logs are designed not to contain prompt content and its policy does not grant general-purpose model-training rights. Its policy permits short operational content retention and longer preservation for support, abuse review, legal compliance, or disputes; its terms also state that end-to-end encryption between consumer and provider is not currently guaranteed.
Data handling can vary by model and deployment. The privacy badge above shows the more conservative of the provider-wide fallback and the policies for routes on this page; each model page shows the policy for that specific route.
Models served by Darkbloom
15 models
- Context
- 131k
- Input /M
- $0.0441/M
- Output /M
- $0.231/M
- Cache read /M
- N/A
- Cache hit
- 0%
- TTFT
- 1.8s
- Throughput
- 21 tps
- Context
- 262k
- Input /M
- $0.0682/M
- Output /M
- $0.189/M
- Cache read /M
- $0.0341/M
- Cache hit
- N/A
- TTFT
- 3.7s
- Throughput
- 24 tps
- Context
- 131k
- Input /M
- $0.0189/M
- Output /M
- $0.0945/M
- Cache read /M
- N/A
- Cache hit
- 0%
- TTFT
- 1.4s
- Throughput
- 50 tps
- Context
- 1.0M
- Input /M
- $0.105/M
- Output /M
- $0.294/M
- Cache read /M
- $0.0021/M
- Cache hit
- N/A
- TTFT
- 3.4s
- Throughput
- 31 tps
Nvidia Nemotron 3.5 Lightning
TEE attested
- Context
- 262k
- Input /M
- $0.0263/M
- Output /M
- $0.126/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- N/A
- Throughput
- N/A
Nvidia Nemotron 3.5 Lightning Thinking
TEE attested
- Context
- 262k
- Input /M
- $0.0263/M
- Output /M
- $0.126/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- N/A
- Throughput
- N/A
- Context
- 262k
- Input /M
- $0.063/M
- Output /M
- $0.367/M
- Cache read /M
- $0.0394/M
- Cache hit
- N/A
- TTFT
- 13.3s
- Throughput
- 40.5 tps
- Context
- 262k
- Input /M
- $0.0525/M
- Output /M
- $0.735/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- 2.7s
- Throughput
- 42 tps
- Context
- 262k
- Input /M
- $0.0682/M
- Output /M
- $0.787/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- 1.3s
- Throughput
- 46 tps
Qwen3.5 9B
FP4
- Context
- 262k
- Input /M
- $0.084/M
- Output /M
- $0.136/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- 2.1s
- Throughput
- 26 tps
- Context
- 262k
- Input /M
- $0.0367/M
- Output /M
- $2.31/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- 2.4s
- Throughput
- 25 tps
- Context
- 262k
- Input /M
- $0.0367/M
- Output /M
- $2.31/M
- Cache read /M
- N/A
- Cache hit
- N/A
- TTFT
- 2.4s
- Throughput
- 25 tps
- Context
- 262k
- Input /M
- $0.0787/M
- Output /M
- $0.525/M
- Cache read /M
- $0.0394/M
- Cache hit
- N/A
- TTFT
- 3.9s
- Throughput
- 16 tps
- Context
- 262k
- Input /M
- $0.084/M
- Output /M
- $0.136/M
- Cache read /M
- $0.042/M
- Cache hit
- N/A
- TTFT
- 1.5s
- Throughput
- 42 tps
Prices are per million tokens. TTFT is the observed time to first token. Selectable routes include the applicable provider-selection markup; fixed routes show the standard model price. Availability and pricing refresh continuously; each model page shows the live provider comparison.
Using Darkbloom via the API
Call a model served by Darkbloom using its displayed model ID, including the provider suffix when selection is supported.
cURL
curl https://nano-gpt.com/api/v1/chat/completions \
-H "Authorization: Bearer $NANOGPT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google/gemma-4-26b-a4b-it:darkbloom",
"messages": [{"role": "user", "content": "Hello"}]
}'For models with provider selection, append :darkbloom to the model ID, set "provider": "darkbloom" in the request body, or send an X-Provider: darkbloom header. Explicit provider selection adds a route-specific markup over that provider's base price; the prices in the table above already include it. Models without provider selection use the displayed model ID without a provider suffix or selection surcharge.
See the API documentation for provider preferences, price-aware routing, and error behavior.