| name | consult-ring |
| description | Consult Ring 2.6 1T (inclusionai/ring-2.6-1t) on OpenRouter for deep reasoning on quant, strategy, hypothesis, or code questions. Use when the user says "/consult-ring261T", "/ask-ring261T", "ask ring", "ring second opinion", or wants the 1T-parameter extended-thinking model's take on a question. |
/consult-ring261T — Ring 2.6 1T Consultation
Consult Ring 2.6 1T (inclusionai/ring-2.6-1t) — a 1-trillion-parameter reasoning model with 262K context, available via OpenRouter. Specialises in chain-of-thought analysis, hypothesis evaluation, and quantitative strategy critique.
Aliases: /consult-ring261T, /ask-ring261T, "ask ring", "ring second opinion"
Prerequisites
OPENROUTER_API_KEY set in Windows environment (key: sk-or-v1-c05470f494...).
- Model:
inclusionai/ring-2.6-1t (hard-coded; do NOT override with OPENROUTER_MODEL env var — other openrouter calls use different models).
- Endpoint:
https://openrouter.ai/api/v1/chat/completions
Usage
/consult-ring261T <question or prompt text>
/ask-ring261T <question or prompt text>
How to invoke
Recommended: python inline (no prompt-file needed)
Write the prompt to a temp file, then call api_consult.py with OPENROUTER_MODEL forced to inclusionai/ring-2.6-1t:
# Write prompt to temp file
$prompt = "Your question here"
$prompt | Out-File -FilePath "$env:TEMP\ring_prompt.txt" -Encoding utf8
# Call via api_consult.py with model override
$env:OPENROUTER_MODEL = "inclusionai/ring-2.6-1t"
$env:OPENROUTER_MAX_TOKENS = "2000"
python tools/swarm/api_consult.py --provider openrouter --prompt-file "$env:TEMP\ring_prompt.txt"
Bash version (via Bash tool)
PROMPT="Your question here"
echo "$PROMPT" > /tmp/ring_prompt.txt
OPENROUTER_MODEL="inclusionai/ring-2.6-1t" OPENROUTER_MAX_TOKENS=2000 \
python tools/swarm/api_consult.py --provider openrouter --prompt-file /tmp/ring_prompt.txt
Direct Python (when api_consult.py is unavailable or for one-liners)
import os, json, urllib.request
key = os.environ["OPENROUTER_API_KEY"]
data = json.dumps({
"model": "inclusionai/ring-2.6-1t",
"messages": [{"role": "user", "content": PROMPT}],
"max_tokens": 2000,
"temperature": 0.3
}).encode()
req = urllib.request.Request(
"https://openrouter.ai/api/v1/chat/completions",
data=data,
headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"}
)
with urllib.request.urlopen(req, timeout=120) as r:
resp = json.loads(r.read())
print(resp["choices"][0]["message"]["content"])
Required parameters
| Parameter | Value |
|---|
model | inclusionai/ring-2.6-1t (ALWAYS — never use default openrouter model) |
max_tokens | 2000 minimum — Ring returns content=null + dumps answer in reasoning_details when tokens are too low. Use 2000 for short questions, 4000 for deep analysis. |
temperature | 0.3 (good for analysis); use 0.7 for creative/brainstorm |
timeout | 120s — Ring's 1T parameter count means 30-90s response time is normal |
Critical gotchas
content=null bug: If max_tokens < 1500, Ring puts the real answer in choices[0].message.reasoning_details and leaves content=null. Always use max_tokens >= 2000.
- OPENROUTER_MODEL env var conflict: Other OpenRouter calls in this repo use different models. Set
OPENROUTER_MODEL=inclusionai/ring-2.6-1t only for Ring calls, then unset or reset afterwards.
- Reasoning trace: Ring's
reasoning_details field contains chain-of-thought reasoning (often very long). The final answer is always in choices[0].message.content when max_tokens >= 2000.
- Response time: 30-90 seconds is normal for this model. Do not retry on timeout until 120s has elapsed.
When to use Ring vs other models
| Use Ring when... | Use something else when... |
|---|
| Deep hypothesis evaluation (H-0xx series) | Quick yes/no gate check (use swarm deepseek) |
| Strategy autopsy with long context | Fast iteration / many small questions (use ofox/groq) |
| Multi-factor quant synthesis (COMMODITY/FOREX/CRYPTO) | Code generation tasks (use kilo/cursor/codex) |
| Frontier-level critique of gate logic | Routine summarisation (use any cheap model) |
| 262K context is needed (large log files, full reports) | Cost matters — Ring tokens are expensive |
What Ring is good at (observed in this repo)
- Evaluating hypothesis registry entries (H-001 through H-037+)
- Ranking free-data vs paid-data hypotheses
- Critiquing gate logic (MDD, DSR, PBO, WR floors)
- Recommending asset class priority given PF/WR/n data
- Synthesising multi-session audit outputs into actionable decisions
Verify alive
OPENROUTER_MODEL="inclusionai/ring-2.6-1t" OPENROUTER_MAX_TOKENS=2000 \
python tools/swarm/api_consult.py --provider openrouter \
--prompt-file <(echo "Reply with exactly the word: VERIFIED")
Expect a response containing VERIFIED.
Related
- Sibling skills:
consult-grok, consult-ofox, consult-gemini, consult-kilo
- Session notes: Ring reviewed H-017 (liquidation cascade) and H-037 (VIX carry) on 2026-05-19
- Ring's H-037 recommendation: VIX term-structure carry is the top free-data hypothesis — pre-registered as H-037 in
reports/hypothesis_registry.json
- Ring's COMMODITY verdict: enforce MDD gate (40% cap) — user decision still pending