New Models: GPT-6.1 Sol and Claude Sonnet 5.5

Date: 2026-09-30 / (1405-07-08)

Summary

GPT-6.1 Sol and Claude Sonnet 5.5 are now available through AvalAI from Tier 1. Both support Chat Completions and Messages fully; GPT-6.1 Sol also has full Responses support, while Sonnet 5.5 has partial Responses support. Pricing and migration guidance are below.

Details

OpenAI: GPT-6.1 Sol

Use gpt-6.1-sol for coding, document analysis, and professional workflows that need reasoning at lower token prices than GPT-6 Astra.

• Limits: 922,000 maximum input tokens and 128,000 maximum output tokens; these are separate catalog limits, not additive capacity guarantees.
• Capabilities: text and image input, text output, reasoning, PDF input, function calling, structured output, and prompt caching. Image understanding does not make this an image-generation model.
• Provider-reported improvements: stronger coding and professional-document work than GPT-6 Sol, improved computer-use and science evaluations, and fewer factual errors in difficult test prompts. OpenAI reports matching Astra on DeepSWE v1.1 at roughly one-fifth task cost; this is an upstream evaluation, not an AvalAI latency, cost-per-task, or benchmark guarantee.
• Cache reads: $0.10 per million tokens at the standard tier, 95% below the standard input rate and half GPT-6 Sol's cache-read rate. Input/output prices remain $2.00/$10.00 at that tier.
• Reasoning: start with the minimal request below. Use reasoning_effort in Chat Completions or reasoning.effort in Responses only with values verified for your route; none and minimal are not supported by the local model metadata. Do not inherit an effort default from older GPT models.

Anthropic: Claude Sonnet 5.5

Use claude-sonnet-5-5 for scoped coding tasks, bug fixes, document creation, spreadsheets, slides, and collaborative design work. It complements Opus 5.5 rather than replacing it for the most complex open-ended work.

• Limits: 1,000,000 maximum input tokens and 128,000 maximum output tokens; leave output headroom for reasoning and the final answer.
• Capabilities: reasoning, image and PDF understanding, tools, structured output, and prompt caching.
• Provider-reported efficiency: Anthropic reports more than 30% faster output generation and up to 30% lower task cost than Sonnet 5 through reduced token use. These figures are workload-dependent upstream measurements, not AvalAI service guarantees.
• Effort defaults differ by product: Anthropic documents medium in Claude apps and Claude Code, but high on the Claude Platform. Do not copy Opus 5.5's API default. Select an explicitly supported effort when predictable behavior matters.
• Thinking-off migration: applications that previously disabled thinking must switch to the new between_tools setting before upgrading. It keeps upfront thinking off, not all thinking off. Follow the Sonnet 5.5 migration guide and verify the setting on your AvalAI route instead of copying a legacy disabled-thinking payload.
• Preserved thinking: account-bound preserved thinking is expanded. Keep signed thinking blocks and complete assistant/tool turns intact; review migration requirements before transferring conversations between accounts. See Anthropic's preserved-thinking guide.
• Request restrictions: do not send unsupported sampling controls such as temperature or top_p, assistant prefill, or forced tool use. Prefer adaptive thinking to a copied fixed budget_tokens setting.

Endpoint availability

Model ID	v1/chat/completions	v1/messages	v1/responses
gpt-6.1-sol	Full	Full	Full
claude-sonnet-5-5	Full	Full	Partial

Partial Responses support does not promise complete hosted-tool, stored-state, or background-workflow parity. For Sonnet-specific thinking controls and conversation blocks, prefer Messages and test the features you use. Endpoint availability alone does not establish access to provider Batch, Flex, Priority, Fast mode, computer-use tools, or other hosted tools. Existing model IDs are not automatically replaced by this announcement.

Pricing

All prices are USD per 1 million tokens and are AvalAI rates, not inferred upstream cache-lifetime prices.

GPT-6.1 Sol

Total request input length	Input	Cached input	Cache creation input	Output
Up to and including 272K tokens	$2.00	$0.10	$2.50	$10.00
Above 272K tokens	$4.00	$0.20	$5.00	$15.00

Exactly 272K input tokens remains in the lower tier. The total request input length selects the applicable tier, including the output rate; this is not a marginal surcharge on only the tokens above 272K.

Claude Sonnet 5.5

Input	Cached input	Cache creation input	Output
$2.00	$0.20	$4.00	$10.00

AvalAI cache creation is $4.00, not the $2.50 upstream cache-write price quoted in the provider announcement. Do not infer a five-minute or one-hour cache lifetime from this rate. No separate long-context price tier is listed for this model. Check the pricing catalog for account-tier throughput limits.

Chat Completions request and SDK examples

Set AVALAI_API_KEY in your environment. Use the same prompt and request shape with either model; change only the model ID when evaluating Sonnet. The examples deliberately omit sampling and effort overrides.

Bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6.1-sol",
    "messages": [{"role": "user", "content": "Review this release plan and give a concise verification checklist: deploy an API behind a feature flag."}]
  }'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"], base_url="https://api.avalai.ir/v1"
)
for model in ("gpt-6.1-sol", "claude-sonnet-5-5"):
    response = client.chat.completions.create(
        model=model,
        messages=[
            {
                "role": "user",
                "content": "Review this release plan and give a concise verification checklist: deploy an API behind a feature flag.",
            }
        ],
    )
    print(model, response.choices[0].message.content)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({ apiKey: process.env.AVALAI_API_KEY, baseURL: "https://api.avalai.ir/v1" });
for (const model of ["gpt-6.1-sol", "claude-sonnet-5-5"]) {
  const response = await client.chat.completions.create({
    model,
    messages: [{ role: "user", content: "Review this release plan and give a concise verification checklist: deploy an API behind a feature flag." }],
  });
  console.log(model, response.choices[0].message.content);
}

Illustrative responses

These are illustrative, not captured live API responses. Each uses 100 uncached prompt tokens and 50 billed completion tokens, no tools, and an example exchange rate of 100,000 IRT/USD. Actual token usage, reasoning-token details, identifiers, and account exchange rates vary. The example cost is (100 × $2 + 50 × $10) / 1,000,000 = $0.0007, or 70 IRT at that example rate.

{
  "id": "chatcmpl-gpt-6-1-sol-example",
  "created": 1790726400,
  "object": "chat.completion",
  "model": "gpt-6.1-sol",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "role": "assistant",
        "content": "Test both flag states, verify authentication and error handling, monitor the limited rollout, and confirm rollback restores the previous behavior."
      }
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150,
    "prompt_tokens_details": {
      "cached_tokens": 0
    }
  },
  "estimated_cost": {
    "unit": "0.0007000000",
    "irt": 70,
    "exchange_rate": 100000
  }
}

{
  "id": "chatcmpl-claude-sonnet-5-5-example",
  "created": 1790726400,
  "object": "chat.completion",
  "model": "claude-sonnet-5-5",
  "choices": [
    {
      "index": 0,
      "finish_reason": "stop",
      "message": {
        "role": "assistant",
        "content": "Test both flag states, verify authentication and error handling, monitor the limited rollout, and confirm rollback restores the previous behavior."
      }
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150,
    "prompt_tokens_details": {
      "cached_tokens": 0
    }
  },
  "estimated_cost": {
    "unit": "0.0007000000",
    "irt": 70,
    "exchange_rate": 100000
  }
}

GPT-6.1 Sol with Responses

This minimal stateless request uses GPT-6.1 Sol's full Responses endpoint. Test any additional hosted-tool or stateful behavior separately.

Bash

curl https://api.avalai.ir/v1/responses \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-6.1-sol","input":"Give a concise verification checklist for a feature-flagged API release."}'

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"], base_url="https://api.avalai.ir/v1"
)
response = client.responses.create(
    model="gpt-6.1-sol",
    input="Give a concise verification checklist for a feature-flagged API release.",
)
print(response.output_text)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({ apiKey: process.env.AVALAI_API_KEY, baseURL: "https://api.avalai.ir/v1" });
const response = await client.responses.create({
  model: "gpt-6.1-sol",
  input: "Give a concise verification checklist for a feature-flagged API release.",
});
console.log(response.output_text);

Claude Sonnet 5.5 with the native Anthropic SDK

Use the host-only base URL for the Anthropic SDK, not a URL ending in /v1. This example explicitly requests adaptive thinking and high effort; it is not the between_tools migration payload. Increase the output budget for harder tasks, and preserve the complete returned content for subsequent tool turns.

Bash

curl https://api.avalai.ir/v1/messages \
  -H "x-api-key: $AVALAI_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model":"claude-sonnet-5-5",
    "max_tokens":4096,
    "thinking":{"type":"adaptive"},
    "output_config":{"effort":"high"},
    "messages":[{"role":"user","content":"Review this release plan and give a concise verification checklist: deploy an API behind a feature flag."}]
  }'

Python

import os
from anthropic import Anthropic

client = Anthropic(
    api_key=os.environ["AVALAI_API_KEY"], base_url="https://api.avalai.ir"
)
response = client.messages.create(
    model="claude-sonnet-5-5",
    max_tokens=4096,
    thinking={"type": "adaptive"},
    output_config={"effort": "high"},
    messages=[
        {
            "role": "user",
            "content": "Review this release plan and give a concise verification checklist: deploy an API behind a feature flag.",
        }
    ],
)
for block in response.content:
    if block.type == "text":
        print(block.text)

JavaScript

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ apiKey: process.env.AVALAI_API_KEY, baseURL: "https://api.avalai.ir" });
const response = await client.messages.create({
  model: "claude-sonnet-5-5",
  max_tokens: 4096,
  thinking: { type: "adaptive" },
  output_config: { effort: "high" },
  messages: [{ role: "user", content: "Review this release plan and give a concise verification checklist: deploy an API behind a feature flag." }],
});
for (const block of response.content) {
  if (block.type === "text") console.log(block.text);
}

Integration checklist

1. Select the explicit model ID and verify Tier 1 access; this announcement does not create aliases or redirects.
2. Re-test output quality, tool loops, schemas, latency, and token usage on your own tasks before switching production traffic.
3. Remove unsupported legacy controls. For Sonnet thinking-off workflows, follow the between_tools migration and account-bound preserved-thinking requirements.
4. Keep complete assistant/tool context, allow enough billed output for reasoning, and request conclusions or concise rationales rather than hidden chain-of-thought.
5. Monitor cache hits and actual billed costs; GPT-6.1 Sol's 272K threshold and Sonnet's AvalAI cache-creation rate can affect estimates.

Related documentation

• OpenAI models and GPT-6.1 Sol reference
• Anthropic models and Sonnet 5.5 reference
• Chat Completions, Messages, and Responses
• Reasoning guide and pricing