deepseek-v4.1-flash
Deepseek V4.1 Flash · Open model, served on Fikra infrastructure
$0.045 in / $0.09 out
per 1M tokens
Model Catalog
One endpoint. Pick a row and call it.
deepseek-v4.1-flash-1017
DeepSeek V4.1 Flash 1017 · The latest DeepSeek V4.1 Flash update (Oct 17 build): better reasoning
$0.05 in / $0.11 out
per 1M tokens
deepseek-v4-flash-0731
Deepseek V4.1 Flash 0731 · Open model, served on Fikra infrastructure
$0.045 in / $0.1 out
per 1M tokens
glm-5.3-flash
GLM 5.3 Flash · Open model, served on Fikra infrastructure
$0.155 in / $0.49 out
per 1M tokens
kimi-k3
Kimi K3 · Open model, served on Fikra infrastructure
$2.1 in / $10.5 out
per 1M tokens
step-5-preview
Step 5 Preview · Open model, served on Fikra infrastructure
$1.25 in / $3.3 out
per 1M tokens
mimo-v2.6-pro
MiMo V2.6 Pro · Xiaomi MiMo-V2.6-Pro: the top-scoring open-weight model on the Artific
$0.5 in / $1.0 out
per 1M tokens
gemini-4-argon
Gemini 4 Argon · Open model, served on Fikra infrastructure
$2.5 in / $12.0 out
per 1M tokens
glm-5.2
GLM 5.2 · Z.ai GLM-5.2 on a 1M context: strong coding and reasoning at a lower p
$0.75 in / $1.65 out
per 1M tokens
glm-5.3
GLM 5.3 · Z.ai flagship GLM-5.3: large-scale reasoning for complex software engi
$1.05 in / $3.35 out
per 1M tokens
nvidia-nemotron-3.5-super-vl
NVIDIA Nemotron 3.5 Super VL · Open model, served on Fikra infrastructure
$0.5 in / $1.25 out
per 1M tokens
qwen3.8-27b
Qwen 3.8 27B · Fast 27B-class multimodal model with automatic prompt caching. Pay $0.
$0.045 in / $0.35 out
per 1M tokens
fikra-pro-20b
Open model, served on Fikra infrastructure
$0.2 in / $0.2 out
per 1M tokens
fikra-pro-120b
Open model, served on Fikra infrastructure
$0.3 in / $0.3 out
per 1M tokens
fikra-qwen-27b
Open model, served on Fikra infrastructure
$0.045 in / $0.35 out
per 1M tokens
fikra-qwen-max
Open model, served on Fikra infrastructure
$6.0 in / $12.0 out
per 1M tokens
fikra-auto
Open model, served on Fikra infrastructure
$0.045 in / $0.35 out
per 1M tokens
fikra-auto-pro
Open model, served on Fikra infrastructure
$6.0 in / $12.0 out
per 1M tokens
fikra-code-35b
Open model, served on Fikra infrastructure
$0.045 in / $0.33 out
per 1M tokens
OPEN-MODEL INFERENCE YOU CAN AUDIT
The rate card is the bill. Compute the cost before the money moves.
OpenAI-compatible endpoints on open-weight models. Among the cheapest live DeepSeek routes anywhere, cached input from $0.025/M, checkpoint IDs that never silently change, up to 1M context, and published benchmarks for every lane, failures included. Pick a model, get a key, and call it with any OpenAI SDK.
OpenAI-compatible
$1 signup credit, no card:
x402, MCP and A2A built in
Try Fikra live
Deploy a bot in one click.
No signup required. Test our infrastructure instantly.
INTEGRATION
Drop-in replacement.
Fikra API is fully OpenAI-compatible. Any SDK, any language. Point your base URL to our endpoint, use your Fikra key, and everything else stays exactly the same.
Read the API Documentation →import openai # Change just this line client = openai.OpenAI( base_url="https://api.fikraapi.co.ke/v1", api_key="your_fikra_key" ) response = client.chat.completions.create( model="deepseek-v4.1-flash-1017", messages=[{ "role": "user", "content": "Analyze this dataset." }] )
WHY FIKRA
What this endpoint actually gives you.
Reasons you should consider Fikra API over direct labs or aggregator gateways, in order of who benefits: agents, agents' operators, and people.
01 PRICE
Cheapest DeepSeek routes in the world
Even cheaper than the official DeepSeek API. The pinned DeepSeek builds are priced at $0.045 to 0.05/M input, with the lowest DeepSeek output rate of any live route, cached input at $0.025/M, and the serving grid transparently engineered for contract-floor price points.
See the DeepSeek line →02 AGENT RAILS
Agents can pay by themselves
x402 (USDC on Base) lets an agent pay per request with no signup, no credit card and no human in the loop. The same /v1 endpoint serves both agent traffic and key traffic, priced at the same rates.
Agent gateway docs →03 PROTOCOLS
MCP and A2A are built in
A first-class MCP server endpoint (fikra_chat, fikra_list_models, fikra_quote) and an A2A message/send route, so agent frameworks can mount Fikra without writing a custom adapter.
Protocol docs →04 COST CONTROL
Deterministic spend, quoted pre-flight
A paid quote endpoint (/v1/agent/quote) returns what a planned call will cost before you run it. Machine-readable pricing.md and a public cost calculator make budgets audit-friendly instead of a surprise line item.
Cost calculator →05 LIFECYCLE
Pinned builds, honest deprecations
Model builds are versioned (deepseek-v4.1-flash-1017, deepseek-v4-flash-0731) and treated as long-lived SKUs. Retirements, when they ever happen, go out on the public feed with a sunset date instead of silently vanishing.
Product feed →06 CACHED INPUT
Cache savings you actually get billed
Prompt caching is automatic on the DeepSeek line: repeated system prompts, RAG context and agent loop prefixes bill at a fraction of fresh input, and the saving shows up in your usage response, not in a rebate you have to claim.
Cached rates →07 TRANSPARENCY
Public evidence, not vibes
Live usage and uptime (usage.json), published traffic rules, a public benchmark page, an OpenAPI spec, llms.txt for agents and a machine-readable model catalog you can audit in a single request.
Usage and uptime →08 GET STARTED
$1 credit to go live
Every new account starts with a $1 signup credit, one top-up scales limits to 3,000 requests/min and 1B tokens/day, and cached input bills from $0.025/M on the DeepSeek line. No subscription to cancel.
Start free →09 BILLING
Money you can compute
Published rates, prepaid wallet, no subscriptions. Normal cards and USDC (x402 on Base) are the rails. Auto-recharge on a threshold if you want it, and the meter stays auditable either way.
Billing details →What's New
Feed
Fikra API
@fikraapi · 2026-10-18
Announcement
Fikra API / Update
✳
step-5-preview scores 8.2/100 on Fikra Eval 500 (preview: unstable)
step-5-preview scored 8.2/100 on Fikra Eval 500. The upstream model is a preview release that degenerates on long prompts (repetition loops) with very high latency. Genuine numbers, unstable model.
→
Fikra API
@fikraapi · 2026-10-18
Announcement
Fikra API / Update
✳
glm-5.3-flash: 80.2/100 on the Fikra Eval 500 benchmark
glm-5.3-flash scored 80.2/100 on Fikra Eval 500.
→
Fikra API
@fikraapi · 2026-10-18
Announcement
Fikra API / Update
✳
Deepseek V4.1 Flash 0731 is now available in Fikra API
Deepseek V4.1 FLah 0731 is a new model from Deepseek with pareto performance improvements
→
Fikra API
@fikraapi · 2026-10-16
New model
Model release
Fikra catalog
Step 5 Preview is now avaiable in Fikra API
Now available through the Fikra API
Step 5 Preview is a new flagship model from Stepfun AI
→PERFORMANCE
The Living Benchmark Suite
Every Fikra model runs the same deterministic 500-task platform evaluation: arithmetic, coding, knowledge, language and instruction following. Results are published automatically the moment a run finishes.
INSIGHTS
Research, News & Blog
COST DETERMINISM
The bill matches the plan.
Published rates, meters you can audit, and no meter surprises. No subscriptions, no platform fees. Top up once and your limits scale permanently. Model usage is billed per token from your prepaid wallet at catalog rates, at dollars you can compute before your agent spends.
Start building: $1 credit, no card →| Limit | New account ($1 signup credit) | After at least 1 top-up |
|---|---|---|
| Requests per minute | 30 | 3,000 |
| Requests per day | 300 | 250,000 |
| Tokens per minute | 100,000 | 20,000,000 |
| Tokens per day | 1,000,000 | 1,000,000,000 |
| Context window | Up to 1,000,000 tokens per request with the DeepSeek line | |
New accounts start with a $1 signup credit, no card required. At $0.045/M input that covers 22M tokens of deepseek-v4.1-flash. The catalog lists live rates for everything else.