pqhaz · Open weights
DeepSeek V4.1 Flash Heretic Thinking
pqhaz's Heretic checkpoint of DeepSeek V4.1 Flash 1017, served by Sable at original mixed precision (FP8 dense/Engram and MXFP4 experts). Supports text and image input, tools, structured output, and optional reasoning. Separate from the DeAlign uncensored checkpoint.
- Reasoning
- Tools
- Vision
- Input
- Text, Images
- Output
- Text
- Reasoning effort
- none · minimal · low · medium · high · xhigh · max
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.20
- Output
- $0.79
- Cache read
- $0.0132
- Cache write
- $0.20
Reference prices as published by the model provider. API example
Specifications
- Context window
- 512K
- Max output
- 128K
Side by side
DeepSeek V4.1 Flash Heretic Thinking compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| DeepSeek V4.1 Flash Heretic ThinkingpqhazThis model | DeepSeek V4 Pro Qwen3.5 9B Uncensored HereticQwenRemove | |
|---|---|---|
| Provider | pqhaz | Qwen |
| Discovery tier | Open weights | Open weights |
| Added | Oct 31, 2026 | Nov 7, 2026 |
| Limits | ||
| Context window | 512KBest | 32.77K |
| Max output | 128KBest | 8.19K |
| Pricing, USD per 1M tokens | ||
| Input | $0.20Best | $0.28 |
| Output | $0.79 | $0.28Best |
| Cache read | $0.0132Best | $0.14 |
| Cache write | $0.20 | — |
| Performance | ||
| Throughput | 186.6 tok/sBest | 68.3 tok/s |
| Time to first token | 2.3 sBest | 86.1 s |
| Request success (24h) | 99.1% | 100.0%Best |
| Capabilities | ||
| Reasoning | Yes | No |
| Tools | Yes | No |
| Vision | Yes | No |
| Reasoning effort | none, minimal, low, medium, high, xhigh, max | — |
Public evaluations
Benchmarks
There are no public benchmark results for these models yet.
Live routing data
Performance and reliability
Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.
- Throughput
- 186.6 tok/s
- Faster than 75% of 52 models
- Time to first token
- 2.3 s
- Quicker than 80% of 52 models
- Prompt cache hits
- 97.4%
- About $0.018 per 1M input tokens with caching
- Request success (24h)
- 99.1%
- Reliable
Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.
- Healthy (98% or more)
- Degraded (90–98%)
- Down (under 90%)
- No requests
Developers
Use it through the API
Ophiris’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
curl https://www.ophiris.xyz/api/v1/chat/completions \ -H "Authorization: Bearer $OPHIRIS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "pqhaz/DeepSeek-V4.1-Flash-1017-heretic:thinking", "reasoning_effort": "medium", "messages": [{ "role": "user", "content": "Explain this stack trace." }] }'import osfrom openai import OpenAI client = OpenAI(base_url="https://www.ophiris.xyz/api/v1", api_key=os.environ["OPHIRIS_API_KEY"])reply = client.chat.completions.create( model="pqhaz/DeepSeek-V4.1-Flash-1017-heretic:thinking", messages=[{"role": "user", "content": "Explain this stack trace."}],)print(reply.choices[0].message.content)pqhaz
Related models
Other pqhaz models, starting with the same family.