GPT-5.6 Sol Pricing Cut by 50% on OpenRouter
TL;DR Highlight
GPT-5.6 Sol lists at $4/$20 on OpenRouter, while OpenAI's official price reportedly hasn't changed.
Who Should Read
Backend developers comparing models and provider routes to cut LLM API costs
Core Mechanics
- GPT-5.6 Sol price and specs
- OpenRouter lists $4 input and $20 output per 1M tokens.
- Context is 1.1M tokens and max output is 128,000 tokens.
- Reasoning effort changes scores a lot
- Artificial Analysis Intelligence Index is 33.5 at Low, 44.0 at Xhigh, and 47.0 at Max.
- Non-reasoning scores 28.3.
- Price differs by provider route
- Standard OpenAI and Azure routes are $4/$20 with $0.40 cache reads.
- Flex is $2/$10 and Fast is $8/$40.
- Watch cache write billing
- Prompts of 1024+ tokens are cached automatically and a cache write is billed each time.
- Setting prompt_cache_options.mode to "explicit" turns it off.
- Tool calling and structured outputs work
- tools and tool_choice enable function calling.
Community Reactions
- Official price reportedly unchanged
- Commenters said the cut is OpenRouter-only and the title is misleading.
- Several noted OpenAI's docs still show the undiscounted price.
- ZDR users may not benefit
- One commenter guessed the discount applies only to the non-ZDR OpenAI route, with Azure at the old price.
- Users with a global ZDR-only toggle may miss out.
- Price competition from Chinese open models
- One team's evals found Kimi K3 matched or beat Opus and Sol in their use cases.
- All their developers switched to K3, and they said the new Sol price is close to K3's.
- Real-world feedback is mixed
- One user liked Sol 5.6 for using fewer tokens and less thinking time and is considering dropping Claude.
- Another reported a $680 coding session that damaged the codebase.
How to Apply
- Price your traffic per route
- Pull one day's input, output and cache tokens and multiply by $4/$20 and Flex $2/$10.
- Record the monthly difference in a table.
- A/B cache write billing
- Send prompt_cache_options.mode as "explicit" on requests with 1024+ token prompts that are rarely reused.
- Compare billed cost before and after.
- Check routing limits if you need ZDR
- Turn on ZDR-only in OpenRouter and verify on your invoice whether the discounted route is excluded.
- If so, budget at $4.40/$22 routes like Azure.
- Split reasoning effort by task
- Send simple requests at Low or Medium and only hard tasks at Xhigh.
- Log score and cost together on the same eval set.
Try It
# Call GPT-5.6 Sol via OpenRouter and disable cache write billing (per the source FAQ; run in a terminal)
# pip install openai
import os
from openai import OpenAI
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=os.environ["OPENROUTER_API_KEY"])
r = client.chat.completions.create(
model="openai/gpt-5.6-sol",
messages=[{"role": "user", "content": "{{question}}"}],
extra_body={"prompt_cache_options": {"mode": "explicit"}},
)
print(r.choices[0].message.content)
print(r.usage)
# Next: remove the extra_body line, resend the same request, and compare usage and billed costTerminology
Related Papers
Mercury 3: Diffusion at scale
Scaling diffusion LLM 40x gives Mercury 3: 4–8x faster than speed-tier models, similar quality
Timestral: Mistral's time-series foundation models
A 209M time-series model ranks first on three benchmarks and runs on CPU.
DeepSeek V4.1 Flash 1017 (Max)
DeepSeek V4.1 Flash 1017 Max scores 45 on the Intelligence Index at 211 tok/s and $0.30 input.
If you don't run your own LLM proxy, how do you know whats going on?
Running your own LLM proxy shows which keys, models, and cache misses drive your usage.
DeepSeek: DeepSeek v4.1 Flash 1017
DeepSeek V4.1 Flash 1017: $0.05/M input, 1M context, built for agents
DeepSeek-v4.1-Flash-1017
DeepSeek V4.1 Flash minor update lifts coding and agent benchmarks while cutting run time and output length
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware