Every number on this page is read from the same model registry that answers GET /v1/models. Measured reliability, latency and cache hit rates: public stats. Prices are quoted in USD per 1M tokens. Card and mobile money top-ups are both charged in KES at KES 133 = $1 and credit the USD wallet. Agents can pay per call in USDC on Base with x402.
Input, cached input and output are itemised. Cached input never costs more than fresh input.
| Model | Family | Input / 1M | Cached / 1M | Output / 1M | Context | |
|---|---|---|---|---|---|---|
| Deepseek V4.1 Flash deepseek-v4.1-flash |
deepseek | $0.05 | $0.0075 | $0.15 | 1.0M | Details |
| DeepSeek V4.1 Flash 1017 deepseek-v4.1-flash-1017 |
deepseek | $0.05 | $0.0075 | $0.15 | 1.0M | Details |
| Deepseek V4.1 Flash 0731 deepseek-v4-flash-0731 |
deepseek | $0.05625 | $0.01125 | $0.17 | 1.0M | Details |
| GLM 5.3 Flash glm-5.3-flash |
glm | $0.155 | $0.08 | $0.49 | 0.31M | Details |
| Kimi K3 kimi-k3 |
kimi | $2.1 | $0.6 | $10.5 | 1.06M | Details |
| Step 5 Preview step-5-preview |
stepfun | $1.25 | same as input | $3.3 | 1.0M | Details |
| MiMo V2.6 Pro mimo-v2.6-pro |
mimo | $0.5 | $0.012 | $1.0 | 1.0M | Details |
| DeepSeek V4 Flash (ModelArk) deepseek-v4-flash-ark |
deepseek | $0.55 | $0.0175 | $1.65 | 1.0M | Details |
| Gemini 4 Argon gemini-4-argon |
gemini | $2.5 | same as input | $12.0 | 1.0M | Details |
| GLM 5.3 glm-5.3 |
glm | $1.05 | $0.55 | $3.35 | 0.31M | Details |
| GLM 5.2 glm-5.2 |
glm | $0.75 | $0.55 | $1.65 | 0.25M | Details |
| NVIDIA Nemotron 3.5 Super VL nvidia-nemotron-3.5-super-vl |
nvidia | $0.5 | same as input | $1.25 | 1.0M | Details |
| Qwen 3.8 27B qwen3.8-27b |
Qwen | $0.045 | $0.045 | $0.35 | 0.96M | Details |
| DeepSeek V4 Pro (ModelArk) deepseek-v4-pro-ark |
deepseek | $1.65 | $0.055 | $4.95 | 1.0M | Details |
| DeepSeek V4.1 Flash (ModelArk) deepseek-v4.1-flash-ark |
deepseek | $0.1875 | $0.00375 | $0.75 | 1.0M | Details |
| DeepSeek V4.1 Flash Economy — free window deepseek-v4-1-flash-economy-free |
deepseek | $0.0 | same as input | $0.0 | 1.0M | Details |
| fikra-pro-20b fikra-pro-20b |
Fikra | $0.2 | same as input | $0.2 | 0.12M | Details |
| fikra-pro-120b fikra-pro-120b |
Fikra | $0.3 | same as input | $0.3 | 0.12M | Details |
| fikra-qwen-27b fikra-qwen-27b |
Fikra | $0.045 | same as input | $0.35 | 0.25M | Details |
| fikra-qwen-max fikra-qwen-max |
Fikra | $6.0 | same as input | $12.0 | 1.0M | Details |
| fikra-auto fikra-auto |
Fikra | $0.045 | same as input | $0.35 | 0.25M | Details |
| fikra-auto-pro fikra-auto-pro |
Fikra | $6.0 | same as input | $12.0 | 1.0M | Details |
| fikra-code-35b fikra-code-35b |
Fikra | $0.045 | same as input | $0.33 | 0.25M | Details |
Repeated prompt prefixes are billed at the cached rate. Cached input can never cost more than fresh input.
| Model | What it does | Rate |
|---|---|---|
| Fikra Speech-to-Text fikra-stt |
Transcribe supported audio into text via the /v1/audio/transcriptions endpoint. | $0.001 / minute |
| MiniMax H3 Video minimax-h3 |
Text-to-video and image-to-video generation. 4-15 second clips, 1080p, first-frame image support. Billed per finished generation. | $0.5 / generation |
| Fikra Image fikra-image |
Text-to-image generation in the OpenAI Images API shape. Billed per finished image. | $0.01 / generation |
| Fikra Text-to-Speech fikra-tts |
Generate natural speech from text via the /v1/audio/speech endpoint. | $25.0 / 1M characters |
| Fikra Embeddings fikra-embed |
Generate embeddings for semantic search, retrieval and RAG via the /v1/embeddings endpoint. | $0.05 / 1M tokens |
A new account gets $0.5 in credit. A first top-up of at least $1.0 activates paid limits.
Requests deduct from the wallet in real time. There is one published rate per model. No seat fee.
Paid accounts get 3,000 requests/minute, 250k requests/day, no token ceilings. There is no new subscription to cancel.
New accounts are prepaid. Existing subscriptions remain active and are honored; they are managed from the dashboard. Public pricing does not sell new subscriptions.
Legacy plans are managed in the dashboard. Public pricing sells prepaid credit only, because that is what the billing system creates for new accounts.
See the change log ->| Tier | Requests / minute | Requests / day | Tokens / minute | Tokens / day |
|---|---|---|---|---|
| Unpaid | 30 | 300 | 100k | 1M |
| Paid | 3,000 | 250,000 | none | none |
These are the limits enforced in code. Volume accounts can request a custom request quota. Limits docs.
New accounts are prepaid. Existing subscriptions remain active and are honored. They are managed from the dashboard. Public pricing does not sell a new subscription.
USD per 1M tokens keeps the price comparable across the market. Card and mobile money checkouts are charged in KES at KES 133 = $1, because KES processing is faster, cheaper and more reliable locally. The wallet is credited in USD, and the dashboard previews the exact KES charge before you pay. Agents can avoid card settlement entirely with x402 USDC.
Yes, when the model publishes a cached rate. It can never be higher than the fresh input rate. If a provider does not support caching, cached tokens bill at the fresh rate.
Each model page lists the upstream family and the retention note that applies. Fikra routes to third-party inference providers; it does not own GPUs.
No contractual SLA on the self-serve plan. Uptime and incidents are published on the status page. Enterprise terms are separate. See the trust centre.
pricing.md, GET /v1/models and the llms.txt index. The website table reads from the same registry.